xref: /netbsd-src/sys/netinet6/in6.c (revision fdd524d4ccd2bb0c6f67401e938dabf773eb0372)
1 /*	$NetBSD: in6.c,v 1.208 2016/07/08 06:18:29 ozaki-r Exp $	*/
2 /*	$KAME: in6.c,v 1.198 2001/07/18 09:12:38 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1991, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)in.c	8.2 (Berkeley) 11/15/93
62  */
63 
64 #include <sys/cdefs.h>
65 __KERNEL_RCSID(0, "$NetBSD: in6.c,v 1.208 2016/07/08 06:18:29 ozaki-r Exp $");
66 
67 #ifdef _KERNEL_OPT
68 #include "opt_inet.h"
69 #include "opt_compat_netbsd.h"
70 #endif
71 
72 #include <sys/param.h>
73 #include <sys/ioctl.h>
74 #include <sys/errno.h>
75 #include <sys/malloc.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/sockio.h>
79 #include <sys/systm.h>
80 #include <sys/proc.h>
81 #include <sys/time.h>
82 #include <sys/kernel.h>
83 #include <sys/syslog.h>
84 #include <sys/kauth.h>
85 #include <sys/cprng.h>
86 #include <sys/kmem.h>
87 
88 #include <net/if.h>
89 #include <net/if_types.h>
90 #include <net/if_llatbl.h>
91 #include <net/if_ether.h>
92 #include <net/if_dl.h>
93 #include <net/pfil.h>
94 #include <net/route.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_var.h>
98 
99 #include <netinet/ip6.h>
100 #include <netinet6/ip6_var.h>
101 #include <netinet6/nd6.h>
102 #include <netinet6/mld6_var.h>
103 #include <netinet6/ip6_mroute.h>
104 #include <netinet6/in6_ifattach.h>
105 #include <netinet6/scope6_var.h>
106 
107 #include <net/net_osdep.h>
108 
109 #ifdef COMPAT_50
110 #include <compat/netinet6/in6_var.h>
111 #endif
112 
113 MALLOC_DEFINE(M_IP6OPT, "ip6_options", "IPv6 options");
114 
115 /* enable backward compatibility code for obsoleted ioctls */
116 #define COMPAT_IN6IFIOCTL
117 
118 #ifdef	IN6_DEBUG
119 #define	IN6_DPRINTF(__fmt, ...)	printf(__fmt, __VA_ARGS__)
120 #else
121 #define	IN6_DPRINTF(__fmt, ...)	do { } while (/*CONSTCOND*/0)
122 #endif /* IN6_DEBUG */
123 
124 /*
125  * Definitions of some constant IP6 addresses.
126  */
127 const struct in6_addr in6addr_any = IN6ADDR_ANY_INIT;
128 const struct in6_addr in6addr_loopback = IN6ADDR_LOOPBACK_INIT;
129 const struct in6_addr in6addr_nodelocal_allnodes =
130 	IN6ADDR_NODELOCAL_ALLNODES_INIT;
131 const struct in6_addr in6addr_linklocal_allnodes =
132 	IN6ADDR_LINKLOCAL_ALLNODES_INIT;
133 const struct in6_addr in6addr_linklocal_allrouters =
134 	IN6ADDR_LINKLOCAL_ALLROUTERS_INIT;
135 
136 const struct in6_addr in6mask0 = IN6MASK0;
137 const struct in6_addr in6mask32 = IN6MASK32;
138 const struct in6_addr in6mask64 = IN6MASK64;
139 const struct in6_addr in6mask96 = IN6MASK96;
140 const struct in6_addr in6mask128 = IN6MASK128;
141 
142 const struct sockaddr_in6 sa6_any = {sizeof(sa6_any), AF_INET6,
143 				     0, 0, IN6ADDR_ANY_INIT, 0};
144 
145 struct pslist_head	in6_ifaddr_list;
146 
147 static int in6_lifaddr_ioctl(struct socket *, u_long, void *,
148 	struct ifnet *);
149 static int in6_ifinit(struct ifnet *, struct in6_ifaddr *,
150 	const struct sockaddr_in6 *, int);
151 static void in6_unlink_ifa(struct in6_ifaddr *, struct ifnet *);
152 
153 void
154 in6_init(void)
155 {
156 
157 	PSLIST_INIT(&in6_ifaddr_list);
158 }
159 
160 /*
161  * Add ownaddr as loopback rtentry.  We previously add the route only if
162  * necessary (ex. on a p2p link).  However, since we now manage addresses
163  * separately from prefixes, we should always add the route.  We can't
164  * rely on the cloning mechanism from the corresponding interface route
165  * any more.
166  */
167 void
168 in6_ifaddlocal(struct ifaddr *ifa)
169 {
170 
171 	if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &in6addr_any) ||
172 	    (ifa->ifa_ifp->if_flags & IFF_POINTOPOINT &&
173 	    IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), IFA_DSTIN6(ifa))))
174 	{
175 		rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
176 		return;
177 	}
178 
179 	rt_ifa_addlocal(ifa);
180 }
181 
182 /*
183  * Remove loopback rtentry of ownaddr generated by in6_ifaddlocal(),
184  * if it exists.
185  */
186 void
187 in6_ifremlocal(struct ifaddr *ifa)
188 {
189 	struct in6_ifaddr *ia;
190 	struct ifaddr *alt_ifa = NULL;
191 	int ia_count = 0;
192 
193 	/*
194 	 * Some of BSD variants do not remove cloned routes
195 	 * from an interface direct route, when removing the direct route
196 	 * (see comments in net/net_osdep.h).  Even for variants that do remove
197 	 * cloned routes, they could fail to remove the cloned routes when
198 	 * we handle multple addresses that share a common prefix.
199 	 * So, we should remove the route corresponding to the deleted address.
200 	 */
201 
202 	/*
203 	 * Delete the entry only if exactly one ifaddr matches the
204 	 * address, ifa->ifa_addr.
205 	 *
206 	 * If more than one ifaddr matches, replace the ifaddr in
207 	 * the routing table, rt_ifa, with a different ifaddr than
208 	 * the one we are purging, ifa.  It is important to do
209 	 * this, or else the routing table can accumulate dangling
210 	 * pointers rt->rt_ifa->ifa_ifp to destroyed interfaces,
211 	 * which will lead to crashes, later.  (More than one ifaddr
212 	 * can match if we assign the same address to multiple---probably
213 	 * p2p---interfaces.)
214 	 *
215 	 * XXX An old comment at this place said, "we should avoid
216 	 * XXX such a configuration [i.e., interfaces with the same
217 	 * XXX addressed assigned --ed.] in IPv6...".  I do not
218 	 * XXX agree, especially now that I have fixed the dangling
219 	 * XXX ifp-pointers bug.
220 	 */
221 	IN6_ADDRLIST_READER_FOREACH(ia) {
222 		if (!IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ia->ia_addr.sin6_addr))
223 			continue;
224 		if (ia->ia_ifp != ifa->ifa_ifp)
225 			alt_ifa = &ia->ia_ifa;
226 		if (++ia_count > 1 && alt_ifa != NULL)
227 			break;
228 	}
229 
230 	if (ia_count == 0)
231 		return;
232 
233 	rt_ifa_remlocal(ifa, ia_count == 1 ? NULL : alt_ifa);
234 }
235 
236 int
237 in6_mask2len(struct in6_addr *mask, u_char *lim0)
238 {
239 	int x = 0, y;
240 	u_char *lim = lim0, *p;
241 
242 	/* ignore the scope_id part */
243 	if (lim0 == NULL || lim0 - (u_char *)mask > sizeof(*mask))
244 		lim = (u_char *)mask + sizeof(*mask);
245 	for (p = (u_char *)mask; p < lim; x++, p++) {
246 		if (*p != 0xff)
247 			break;
248 	}
249 	y = 0;
250 	if (p < lim) {
251 		for (y = 0; y < NBBY; y++) {
252 			if ((*p & (0x80 >> y)) == 0)
253 				break;
254 		}
255 	}
256 
257 	/*
258 	 * when the limit pointer is given, do a stricter check on the
259 	 * remaining bits.
260 	 */
261 	if (p < lim) {
262 		if (y != 0 && (*p & (0x00ff >> y)) != 0)
263 			return -1;
264 		for (p = p + 1; p < lim; p++)
265 			if (*p != 0)
266 				return -1;
267 	}
268 
269 	return x * NBBY + y;
270 }
271 
272 #define ifa2ia6(ifa)	((struct in6_ifaddr *)(ifa))
273 #define ia62ifa(ia6)	(&((ia6)->ia_ifa))
274 
275 static int
276 in6_control1(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
277 {
278 	struct	in6_ifreq *ifr = (struct in6_ifreq *)data;
279 	struct	in6_ifaddr *ia = NULL;
280 	struct	in6_aliasreq *ifra = (struct in6_aliasreq *)data;
281 	struct sockaddr_in6 *sa6;
282 	int error;
283 
284 	switch (cmd) {
285 	case SIOCAADDRCTL_POLICY:
286 	case SIOCDADDRCTL_POLICY:
287 		/* Privileged. */
288 		return in6_src_ioctl(cmd, data);
289 	/*
290 	 * XXX: Fix me, once we fix SIOCSIFADDR, SIOCIFDSTADDR, etc.
291 	 */
292 	case SIOCSIFADDR:
293 	case SIOCSIFDSTADDR:
294 	case SIOCSIFBRDADDR:
295 	case SIOCSIFNETMASK:
296 		return EOPNOTSUPP;
297 	case SIOCGETSGCNT_IN6:
298 	case SIOCGETMIFCNT_IN6:
299 		return mrt6_ioctl(cmd, data);
300 	case SIOCGIFADDRPREF:
301 	case SIOCSIFADDRPREF:
302 		if (ifp == NULL)
303 			return EINVAL;
304 		return ifaddrpref_ioctl(so, cmd, data, ifp);
305 	}
306 
307 	if (ifp == NULL)
308 		return EOPNOTSUPP;
309 
310 	switch (cmd) {
311 	case SIOCSNDFLUSH_IN6:
312 	case SIOCSPFXFLUSH_IN6:
313 	case SIOCSRTRFLUSH_IN6:
314 	case SIOCSDEFIFACE_IN6:
315 	case SIOCSIFINFO_FLAGS:
316 	case SIOCSIFINFO_IN6:
317 		/* Privileged. */
318 		/* FALLTHROUGH */
319 	case OSIOCGIFINFO_IN6:
320 	case SIOCGIFINFO_IN6:
321 	case SIOCGDRLST_IN6:
322 	case SIOCGPRLST_IN6:
323 	case SIOCGNBRINFO_IN6:
324 	case SIOCGDEFIFACE_IN6:
325 		return nd6_ioctl(cmd, data, ifp);
326 	}
327 
328 	switch (cmd) {
329 	case SIOCSIFPREFIX_IN6:
330 	case SIOCDIFPREFIX_IN6:
331 	case SIOCAIFPREFIX_IN6:
332 	case SIOCCIFPREFIX_IN6:
333 	case SIOCSGIFPREFIX_IN6:
334 	case SIOCGIFPREFIX_IN6:
335 		log(LOG_NOTICE,
336 		    "prefix ioctls are now invalidated. "
337 		    "please use ifconfig.\n");
338 		return EOPNOTSUPP;
339 	}
340 
341 	switch (cmd) {
342 	case SIOCALIFADDR:
343 	case SIOCDLIFADDR:
344 		/* Privileged. */
345 		/* FALLTHROUGH */
346 	case SIOCGLIFADDR:
347 		return in6_lifaddr_ioctl(so, cmd, data, ifp);
348 	}
349 
350 	/*
351 	 * Find address for this interface, if it exists.
352 	 *
353 	 * In netinet code, we have checked ifra_addr in SIOCSIF*ADDR operation
354 	 * only, and used the first interface address as the target of other
355 	 * operations (without checking ifra_addr).  This was because netinet
356 	 * code/API assumed at most 1 interface address per interface.
357 	 * Since IPv6 allows a node to assign multiple addresses
358 	 * on a single interface, we almost always look and check the
359 	 * presence of ifra_addr, and reject invalid ones here.
360 	 * It also decreases duplicated code among SIOC*_IN6 operations.
361 	 */
362 	switch (cmd) {
363 	case SIOCAIFADDR_IN6:
364 #ifdef OSIOCAIFADDR_IN6
365 	case OSIOCAIFADDR_IN6:
366 #endif
367 #ifdef OSIOCSIFPHYADDR_IN6
368 	case OSIOCSIFPHYADDR_IN6:
369 #endif
370 	case SIOCSIFPHYADDR_IN6:
371 		sa6 = &ifra->ifra_addr;
372 		break;
373 	case SIOCSIFADDR_IN6:
374 	case SIOCGIFADDR_IN6:
375 	case SIOCSIFDSTADDR_IN6:
376 	case SIOCSIFNETMASK_IN6:
377 	case SIOCGIFDSTADDR_IN6:
378 	case SIOCGIFNETMASK_IN6:
379 	case SIOCDIFADDR_IN6:
380 	case SIOCGIFPSRCADDR_IN6:
381 	case SIOCGIFPDSTADDR_IN6:
382 	case SIOCGIFAFLAG_IN6:
383 	case SIOCSNDFLUSH_IN6:
384 	case SIOCSPFXFLUSH_IN6:
385 	case SIOCSRTRFLUSH_IN6:
386 	case SIOCGIFALIFETIME_IN6:
387 #ifdef OSIOCGIFALIFETIME_IN6
388 	case OSIOCGIFALIFETIME_IN6:
389 #endif
390 	case SIOCGIFSTAT_IN6:
391 	case SIOCGIFSTAT_ICMP6:
392 		sa6 = &ifr->ifr_addr;
393 		break;
394 	default:
395 		sa6 = NULL;
396 		break;
397 	}
398 	if (sa6 && sa6->sin6_family == AF_INET6) {
399 		if (sa6->sin6_scope_id != 0)
400 			error = sa6_embedscope(sa6, 0);
401 		else
402 			error = in6_setscope(&sa6->sin6_addr, ifp, NULL);
403 		if (error != 0)
404 			return error;
405 		ia = in6ifa_ifpwithaddr(ifp, &sa6->sin6_addr);
406 	} else
407 		ia = NULL;
408 
409 	switch (cmd) {
410 	case SIOCSIFADDR_IN6:
411 	case SIOCSIFDSTADDR_IN6:
412 	case SIOCSIFNETMASK_IN6:
413 		/*
414 		 * Since IPv6 allows a node to assign multiple addresses
415 		 * on a single interface, SIOCSIFxxx ioctls are deprecated.
416 		 */
417 		return EINVAL;
418 
419 	case SIOCDIFADDR_IN6:
420 		/*
421 		 * for IPv4, we look for existing in_ifaddr here to allow
422 		 * "ifconfig if0 delete" to remove the first IPv4 address on
423 		 * the interface.  For IPv6, as the spec allows multiple
424 		 * interface address from the day one, we consider "remove the
425 		 * first one" semantics to be not preferable.
426 		 */
427 		if (ia == NULL)
428 			return EADDRNOTAVAIL;
429 		/* FALLTHROUGH */
430 #ifdef OSIOCAIFADDR_IN6
431 	case OSIOCAIFADDR_IN6:
432 #endif
433 	case SIOCAIFADDR_IN6:
434 		/*
435 		 * We always require users to specify a valid IPv6 address for
436 		 * the corresponding operation.
437 		 */
438 		if (ifra->ifra_addr.sin6_family != AF_INET6 ||
439 		    ifra->ifra_addr.sin6_len != sizeof(struct sockaddr_in6))
440 			return EAFNOSUPPORT;
441 		/* Privileged. */
442 
443 		break;
444 
445 	case SIOCGIFADDR_IN6:
446 		/* This interface is basically deprecated. use SIOCGIFCONF. */
447 		/* FALLTHROUGH */
448 	case SIOCGIFAFLAG_IN6:
449 	case SIOCGIFNETMASK_IN6:
450 	case SIOCGIFDSTADDR_IN6:
451 	case SIOCGIFALIFETIME_IN6:
452 #ifdef OSIOCGIFALIFETIME_IN6
453 	case OSIOCGIFALIFETIME_IN6:
454 #endif
455 		/* must think again about its semantics */
456 		if (ia == NULL)
457 			return EADDRNOTAVAIL;
458 		break;
459 	}
460 
461 	switch (cmd) {
462 
463 	case SIOCGIFADDR_IN6:
464 		ifr->ifr_addr = ia->ia_addr;
465 		if ((error = sa6_recoverscope(&ifr->ifr_addr)) != 0)
466 			return error;
467 		break;
468 
469 	case SIOCGIFDSTADDR_IN6:
470 		if ((ifp->if_flags & IFF_POINTOPOINT) == 0)
471 			return EINVAL;
472 		/*
473 		 * XXX: should we check if ifa_dstaddr is NULL and return
474 		 * an error?
475 		 */
476 		ifr->ifr_dstaddr = ia->ia_dstaddr;
477 		if ((error = sa6_recoverscope(&ifr->ifr_dstaddr)) != 0)
478 			return error;
479 		break;
480 
481 	case SIOCGIFNETMASK_IN6:
482 		ifr->ifr_addr = ia->ia_prefixmask;
483 		break;
484 
485 	case SIOCGIFAFLAG_IN6:
486 		ifr->ifr_ifru.ifru_flags6 = ia->ia6_flags;
487 		break;
488 
489 	case SIOCGIFSTAT_IN6:
490 		if (ifp == NULL)
491 			return EINVAL;
492 		memset(&ifr->ifr_ifru.ifru_stat, 0,
493 		    sizeof(ifr->ifr_ifru.ifru_stat));
494 		ifr->ifr_ifru.ifru_stat =
495 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->in6_ifstat;
496 		break;
497 
498 	case SIOCGIFSTAT_ICMP6:
499 		if (ifp == NULL)
500 			return EINVAL;
501 		memset(&ifr->ifr_ifru.ifru_icmp6stat, 0,
502 		    sizeof(ifr->ifr_ifru.ifru_icmp6stat));
503 		ifr->ifr_ifru.ifru_icmp6stat =
504 		    *((struct in6_ifextra *)ifp->if_afdata[AF_INET6])->icmp6_ifstat;
505 		break;
506 
507 #ifdef OSIOCGIFALIFETIME_IN6
508 	case OSIOCGIFALIFETIME_IN6:
509 #endif
510 	case SIOCGIFALIFETIME_IN6:
511 		ifr->ifr_ifru.ifru_lifetime = ia->ia6_lifetime;
512 		if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
513 			time_t maxexpire;
514 			struct in6_addrlifetime *retlt =
515 			    &ifr->ifr_ifru.ifru_lifetime;
516 
517 			/*
518 			 * XXX: adjust expiration time assuming time_t is
519 			 * signed.
520 			 */
521 			maxexpire = ((time_t)~0) &
522 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
523 			if (ia->ia6_lifetime.ia6t_vltime <
524 			    maxexpire - ia->ia6_updatetime) {
525 				retlt->ia6t_expire = ia->ia6_updatetime +
526 				    ia->ia6_lifetime.ia6t_vltime;
527 				retlt->ia6t_expire = retlt->ia6t_expire ?
528 				    time_mono_to_wall(retlt->ia6t_expire) :
529 				    0;
530 			} else
531 				retlt->ia6t_expire = maxexpire;
532 		}
533 		if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
534 			time_t maxexpire;
535 			struct in6_addrlifetime *retlt =
536 			    &ifr->ifr_ifru.ifru_lifetime;
537 
538 			/*
539 			 * XXX: adjust expiration time assuming time_t is
540 			 * signed.
541 			 */
542 			maxexpire = ((time_t)~0) &
543 			    ~((time_t)1 << ((sizeof(maxexpire) * NBBY) - 1));
544 			if (ia->ia6_lifetime.ia6t_pltime <
545 			    maxexpire - ia->ia6_updatetime) {
546 				retlt->ia6t_preferred = ia->ia6_updatetime +
547 				    ia->ia6_lifetime.ia6t_pltime;
548 				retlt->ia6t_preferred = retlt->ia6t_preferred ?
549 				    time_mono_to_wall(retlt->ia6t_preferred) :
550 				    0;
551 			} else
552 				retlt->ia6t_preferred = maxexpire;
553 		}
554 #ifdef OSIOCFIFALIFETIME_IN6
555 		if (cmd == OSIOCFIFALIFETIME_IN6)
556 			in6_addrlifetime_to_in6_addrlifetime50(
557 			    &ifr->ifru.ifru_lifetime);
558 #endif
559 		break;
560 
561 #ifdef OSIOCAIFADDR_IN6
562 	case OSIOCAIFADDR_IN6:
563 		in6_aliasreq50_to_in6_aliasreq(ifra);
564 		/*FALLTHROUGH*/
565 #endif
566 	case SIOCAIFADDR_IN6:
567 	{
568 		int i;
569 		struct nd_prefixctl prc0;
570 		struct nd_prefix *pr;
571 		struct in6_addrlifetime *lt;
572 
573 		/* reject read-only flags */
574 		if ((ifra->ifra_flags & IN6_IFF_DUPLICATED) != 0 ||
575 		    (ifra->ifra_flags & IN6_IFF_DETACHED) != 0 ||
576 		    (ifra->ifra_flags & IN6_IFF_TENTATIVE) != 0 ||
577 		    (ifra->ifra_flags & IN6_IFF_NODAD) != 0 ||
578 		    (ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0) {
579 			return EINVAL;
580 		}
581 		/*
582 		 * ia6t_expire and ia6t_preferred won't be used for now,
583 		 * so just in case.
584 		 */
585 		lt = &ifra->ifra_lifetime;
586 		if (lt->ia6t_expire != 0)
587 			lt->ia6t_expire = time_wall_to_mono(lt->ia6t_expire);
588 		if (lt->ia6t_preferred != 0)
589 			lt->ia6t_preferred =
590 			    time_wall_to_mono(lt->ia6t_preferred);
591 		/*
592 		 * first, make or update the interface address structure,
593 		 * and link it to the list.
594 		 */
595 		if ((error = in6_update_ifa(ifp, ifra, ia, 0)) != 0)
596 			return error;
597 		if ((ia = in6ifa_ifpwithaddr(ifp, &ifra->ifra_addr.sin6_addr))
598 		    == NULL) {
599 		    	/*
600 			 * this can happen when the user specify the 0 valid
601 			 * lifetime.
602 			 */
603 			break;
604 		}
605 
606 		/*
607 		 * then, make the prefix on-link on the interface.
608 		 * XXX: we'd rather create the prefix before the address, but
609 		 * we need at least one address to install the corresponding
610 		 * interface route, so we configure the address first.
611 		 */
612 
613 		/*
614 		 * convert mask to prefix length (prefixmask has already
615 		 * been validated in in6_update_ifa().
616 		 */
617 		memset(&prc0, 0, sizeof(prc0));
618 		prc0.ndprc_ifp = ifp;
619 		prc0.ndprc_plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
620 		    NULL);
621 		if (prc0.ndprc_plen == 128) {
622 			break;	/* we don't need to install a host route. */
623 		}
624 		prc0.ndprc_prefix = ifra->ifra_addr;
625 		/* apply the mask for safety. */
626 		for (i = 0; i < 4; i++) {
627 			prc0.ndprc_prefix.sin6_addr.s6_addr32[i] &=
628 			    ifra->ifra_prefixmask.sin6_addr.s6_addr32[i];
629 		}
630 		/*
631 		 * XXX: since we don't have an API to set prefix (not address)
632 		 * lifetimes, we just use the same lifetimes as addresses.
633 		 * The (temporarily) installed lifetimes can be overridden by
634 		 * later advertised RAs (when accept_rtadv is non 0), which is
635 		 * an intended behavior.
636 		 */
637 		prc0.ndprc_raf_onlink = 1; /* should be configurable? */
638 		prc0.ndprc_raf_auto =
639 		    ((ifra->ifra_flags & IN6_IFF_AUTOCONF) != 0);
640 		prc0.ndprc_vltime = ifra->ifra_lifetime.ia6t_vltime;
641 		prc0.ndprc_pltime = ifra->ifra_lifetime.ia6t_pltime;
642 
643 		/* add the prefix if not yet. */
644 		if ((pr = nd6_prefix_lookup(&prc0)) == NULL) {
645 			/*
646 			 * nd6_prelist_add will install the corresponding
647 			 * interface route.
648 			 */
649 			if ((error = nd6_prelist_add(&prc0, NULL, &pr)) != 0)
650 				return error;
651 			if (pr == NULL) {
652 				log(LOG_ERR, "nd6_prelist_add succeeded but "
653 				    "no prefix\n");
654 				return EINVAL; /* XXX panic here? */
655 			}
656 		}
657 
658 		/* relate the address to the prefix */
659 		if (ia->ia6_ndpr == NULL) {
660 			ia->ia6_ndpr = pr;
661 			pr->ndpr_refcnt++;
662 
663 			/*
664 			 * If this is the first autoconf address from the
665 			 * prefix, create a temporary address as well
666 			 * (when required).
667 			 */
668 			if ((ia->ia6_flags & IN6_IFF_AUTOCONF) &&
669 			    ip6_use_tempaddr && pr->ndpr_refcnt == 1) {
670 				int e;
671 				if ((e = in6_tmpifadd(ia, 1, 0)) != 0) {
672 					log(LOG_NOTICE, "in6_control: failed "
673 					    "to create a temporary address, "
674 					    "errno=%d\n", e);
675 				}
676 			}
677 		}
678 
679 		/*
680 		 * this might affect the status of autoconfigured addresses,
681 		 * that is, this address might make other addresses detached.
682 		 */
683 		pfxlist_onlink_check();
684 
685 		(void)pfil_run_hooks(if_pfil, (struct mbuf **)SIOCAIFADDR_IN6,
686 		    ifp, PFIL_IFADDR);
687 		break;
688 	}
689 
690 	case SIOCDIFADDR_IN6:
691 	{
692 		struct nd_prefix *pr;
693 
694 		/*
695 		 * If the address being deleted is the only one that owns
696 		 * the corresponding prefix, expire the prefix as well.
697 		 * XXX: theoretically, we don't have to worry about such
698 		 * relationship, since we separate the address management
699 		 * and the prefix management.  We do this, however, to provide
700 		 * as much backward compatibility as possible in terms of
701 		 * the ioctl operation.
702 		 * Note that in6_purgeaddr() will decrement ndpr_refcnt.
703 		 */
704 		pr = ia->ia6_ndpr;
705 		in6_purgeaddr(&ia->ia_ifa);
706 		if (pr && pr->ndpr_refcnt == 0)
707 			prelist_remove(pr);
708 		(void)pfil_run_hooks(if_pfil, (struct mbuf **)SIOCDIFADDR_IN6,
709 		    ifp, PFIL_IFADDR);
710 		break;
711 	}
712 
713 	default:
714 		return ENOTTY;
715 	}
716 
717 	return 0;
718 }
719 
720 int
721 in6_control(struct socket *so, u_long cmd, void *data, struct ifnet *ifp)
722 {
723 	int error, s;
724 
725 	switch (cmd) {
726 	case SIOCSNDFLUSH_IN6:
727 	case SIOCSPFXFLUSH_IN6:
728 	case SIOCSRTRFLUSH_IN6:
729 	case SIOCSDEFIFACE_IN6:
730 	case SIOCSIFINFO_FLAGS:
731 	case SIOCSIFINFO_IN6:
732 
733 	case SIOCALIFADDR:
734 	case SIOCDLIFADDR:
735 
736 	case SIOCDIFADDR_IN6:
737 #ifdef OSIOCAIFADDR_IN6
738 	case OSIOCAIFADDR_IN6:
739 #endif
740 	case SIOCAIFADDR_IN6:
741 
742 	case SIOCAADDRCTL_POLICY:
743 	case SIOCDADDRCTL_POLICY:
744 
745 		if (kauth_authorize_network(curlwp->l_cred,
746 		    KAUTH_NETWORK_SOCKET,
747 		    KAUTH_REQ_NETWORK_SOCKET_SETPRIV,
748 		    so, NULL, NULL))
749 			return EPERM;
750 		break;
751 	}
752 
753 	s = splnet();
754 	error = in6_control1(so , cmd, data, ifp);
755 	splx(s);
756 	return error;
757 }
758 
759 /*
760  * Update parameters of an IPv6 interface address.
761  * If necessary, a new entry is created and linked into address chains.
762  * This function is separated from in6_control().
763  * XXX: should this be performed under splnet()?
764  */
765 static int
766 in6_update_ifa1(struct ifnet *ifp, struct in6_aliasreq *ifra,
767     struct in6_ifaddr *ia, int flags)
768 {
769 	int error = 0, hostIsNew = 0, plen = -1;
770 	struct sockaddr_in6 dst6;
771 	struct in6_addrlifetime *lt;
772 	struct in6_multi_mship *imm;
773 	struct in6_multi *in6m_sol;
774 	struct rtentry *rt;
775 	int dad_delay, was_tentative;
776 
777 	in6m_sol = NULL;
778 
779 	/* Validate parameters */
780 	if (ifp == NULL || ifra == NULL) /* this maybe redundant */
781 		return EINVAL;
782 
783 	/*
784 	 * The destination address for a p2p link must have a family
785 	 * of AF_UNSPEC or AF_INET6.
786 	 */
787 	if ((ifp->if_flags & IFF_POINTOPOINT) != 0 &&
788 	    ifra->ifra_dstaddr.sin6_family != AF_INET6 &&
789 	    ifra->ifra_dstaddr.sin6_family != AF_UNSPEC)
790 		return EAFNOSUPPORT;
791 	/*
792 	 * validate ifra_prefixmask.  don't check sin6_family, netmask
793 	 * does not carry fields other than sin6_len.
794 	 */
795 	if (ifra->ifra_prefixmask.sin6_len > sizeof(struct sockaddr_in6))
796 		return EINVAL;
797 	/*
798 	 * Because the IPv6 address architecture is classless, we require
799 	 * users to specify a (non 0) prefix length (mask) for a new address.
800 	 * We also require the prefix (when specified) mask is valid, and thus
801 	 * reject a non-consecutive mask.
802 	 */
803 	if (ia == NULL && ifra->ifra_prefixmask.sin6_len == 0)
804 		return EINVAL;
805 	if (ifra->ifra_prefixmask.sin6_len != 0) {
806 		plen = in6_mask2len(&ifra->ifra_prefixmask.sin6_addr,
807 		    (u_char *)&ifra->ifra_prefixmask +
808 		    ifra->ifra_prefixmask.sin6_len);
809 		if (plen <= 0)
810 			return EINVAL;
811 	} else {
812 		/*
813 		 * In this case, ia must not be NULL.  We just use its prefix
814 		 * length.
815 		 */
816 		plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
817 	}
818 	/*
819 	 * If the destination address on a p2p interface is specified,
820 	 * and the address is a scoped one, validate/set the scope
821 	 * zone identifier.
822 	 */
823 	dst6 = ifra->ifra_dstaddr;
824 	if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) != 0 &&
825 	    (dst6.sin6_family == AF_INET6)) {
826 		struct in6_addr in6_tmp;
827 		u_int32_t zoneid;
828 
829 		in6_tmp = dst6.sin6_addr;
830 		if (in6_setscope(&in6_tmp, ifp, &zoneid))
831 			return EINVAL; /* XXX: should be impossible */
832 
833 		if (dst6.sin6_scope_id != 0) {
834 			if (dst6.sin6_scope_id != zoneid)
835 				return EINVAL;
836 		} else		/* user omit to specify the ID. */
837 			dst6.sin6_scope_id = zoneid;
838 
839 		/* convert into the internal form */
840 		if (sa6_embedscope(&dst6, 0))
841 			return EINVAL; /* XXX: should be impossible */
842 	}
843 	/*
844 	 * The destination address can be specified only for a p2p or a
845 	 * loopback interface.  If specified, the corresponding prefix length
846 	 * must be 128.
847 	 */
848 	if (ifra->ifra_dstaddr.sin6_family == AF_INET6) {
849 #ifdef FORCE_P2PPLEN
850 		int i;
851 #endif
852 
853 		if ((ifp->if_flags & (IFF_POINTOPOINT|IFF_LOOPBACK)) == 0) {
854 			/* XXX: noisy message */
855 			nd6log(LOG_INFO, "a destination can "
856 			    "be specified for a p2p or a loopback IF only\n");
857 			return EINVAL;
858 		}
859 		if (plen != 128) {
860 			nd6log(LOG_INFO, "prefixlen should "
861 			    "be 128 when dstaddr is specified\n");
862 #ifdef FORCE_P2PPLEN
863 			/*
864 			 * To be compatible with old configurations,
865 			 * such as ifconfig gif0 inet6 2001::1 2001::2
866 			 * prefixlen 126, we override the specified
867 			 * prefixmask as if the prefix length was 128.
868 			 */
869 			ifra->ifra_prefixmask.sin6_len =
870 			    sizeof(struct sockaddr_in6);
871 			for (i = 0; i < 4; i++)
872 				ifra->ifra_prefixmask.sin6_addr.s6_addr32[i] =
873 				    0xffffffff;
874 			plen = 128;
875 #else
876 			return EINVAL;
877 #endif
878 		}
879 	}
880 	/* lifetime consistency check */
881 	lt = &ifra->ifra_lifetime;
882 	if (lt->ia6t_pltime > lt->ia6t_vltime)
883 		return EINVAL;
884 	if (lt->ia6t_vltime == 0) {
885 		/*
886 		 * the following log might be noisy, but this is a typical
887 		 * configuration mistake or a tool's bug.
888 		 */
889 		nd6log(LOG_INFO, "valid lifetime is 0 for %s\n",
890 		    ip6_sprintf(&ifra->ifra_addr.sin6_addr));
891 
892 		if (ia == NULL)
893 			return 0; /* there's nothing to do */
894 	}
895 
896 	/*
897 	 * If this is a new address, allocate a new ifaddr and link it
898 	 * into chains.
899 	 */
900 	if (ia == NULL) {
901 		hostIsNew = 1;
902 		/*
903 		 * When in6_update_ifa() is called in a process of a received
904 		 * RA, it is called under an interrupt context.  So, we should
905 		 * call malloc with M_NOWAIT.
906 		 */
907 		ia = (struct in6_ifaddr *) malloc(sizeof(*ia), M_IFADDR,
908 		    M_NOWAIT);
909 		if (ia == NULL)
910 			return ENOBUFS;
911 		memset(ia, 0, sizeof(*ia));
912 		LIST_INIT(&ia->ia6_memberships);
913 		/* Initialize the address and masks, and put time stamp */
914 		ia->ia_ifa.ifa_addr = (struct sockaddr *)&ia->ia_addr;
915 		ia->ia_addr.sin6_family = AF_INET6;
916 		ia->ia_addr.sin6_len = sizeof(ia->ia_addr);
917 		ia->ia6_createtime = time_uptime;
918 		if ((ifp->if_flags & (IFF_POINTOPOINT | IFF_LOOPBACK)) != 0) {
919 			/*
920 			 * XXX: some functions expect that ifa_dstaddr is not
921 			 * NULL for p2p interfaces.
922 			 */
923 			ia->ia_ifa.ifa_dstaddr =
924 			    (struct sockaddr *)&ia->ia_dstaddr;
925 		} else {
926 			ia->ia_ifa.ifa_dstaddr = NULL;
927 		}
928 		ia->ia_ifa.ifa_netmask =
929 		    (struct sockaddr *)&ia->ia_prefixmask;
930 
931 		ia->ia_ifp = ifp;
932 		IN6_ADDRLIST_ENTRY_INIT(ia);
933 	}
934 
935 	/* update timestamp */
936 	ia->ia6_updatetime = time_uptime;
937 
938 	/* set prefix mask */
939 	if (ifra->ifra_prefixmask.sin6_len) {
940 		/*
941 		 * We prohibit changing the prefix length of an existing
942 		 * address, because
943 		 * + such an operation should be rare in IPv6, and
944 		 * + the operation would confuse prefix management.
945 		 */
946 		if (ia->ia_prefixmask.sin6_len &&
947 		    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL) != plen) {
948 			nd6log(LOG_INFO, "the prefix length of an"
949 			    " existing (%s) address should not be changed\n",
950 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
951 			error = EINVAL;
952 			if (hostIsNew)
953 				free(ia, M_IFADDR);
954 			goto exit;
955 		}
956 		ia->ia_prefixmask = ifra->ifra_prefixmask;
957 	}
958 
959 	/*
960 	 * If a new destination address is specified, scrub the old one and
961 	 * install the new destination.  Note that the interface must be
962 	 * p2p or loopback (see the check above.)
963 	 */
964 	if (dst6.sin6_family == AF_INET6 &&
965 	    !IN6_ARE_ADDR_EQUAL(&dst6.sin6_addr, &ia->ia_dstaddr.sin6_addr)) {
966 		if ((ia->ia_flags & IFA_ROUTE) != 0 &&
967 		    rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST) != 0) {
968 			nd6log(LOG_ERR, "failed to remove "
969 			    "a route to the old destination: %s\n",
970 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
971 			/* proceed anyway... */
972 		} else
973 			ia->ia_flags &= ~IFA_ROUTE;
974 		ia->ia_dstaddr = dst6;
975 	}
976 
977 	/*
978 	 * Set lifetimes.  We do not refer to ia6t_expire and ia6t_preferred
979 	 * to see if the address is deprecated or invalidated, but initialize
980 	 * these members for applications.
981 	 */
982 	ia->ia6_lifetime = ifra->ifra_lifetime;
983 	if (ia->ia6_lifetime.ia6t_vltime != ND6_INFINITE_LIFETIME) {
984 		ia->ia6_lifetime.ia6t_expire =
985 		    time_uptime + ia->ia6_lifetime.ia6t_vltime;
986 	} else
987 		ia->ia6_lifetime.ia6t_expire = 0;
988 	if (ia->ia6_lifetime.ia6t_pltime != ND6_INFINITE_LIFETIME) {
989 		ia->ia6_lifetime.ia6t_preferred =
990 		    time_uptime + ia->ia6_lifetime.ia6t_pltime;
991 	} else
992 		ia->ia6_lifetime.ia6t_preferred = 0;
993 
994 	/*
995 	 * configure address flags.
996 	 * We need to preserve tentative state so DAD works if
997 	 * something adds the same address before DAD finishes.
998 	 */
999 	was_tentative = ia->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED);
1000 	ia->ia6_flags = ifra->ifra_flags;
1001 
1002 	/*
1003 	 * Make the address tentative before joining multicast addresses,
1004 	 * so that corresponding MLD responses would not have a tentative
1005 	 * source address.
1006 	 */
1007 	ia->ia6_flags &= ~IN6_IFF_DUPLICATED;	/* safety */
1008 	if (ifp->if_link_state == LINK_STATE_DOWN) {
1009 		ia->ia6_flags |= IN6_IFF_DETACHED;
1010 		ia->ia6_flags &= ~IN6_IFF_TENTATIVE;
1011 	} else if ((hostIsNew || was_tentative) && if_do_dad(ifp))
1012 		ia->ia6_flags |= IN6_IFF_TENTATIVE;
1013 
1014 	/*
1015 	 * backward compatibility - if IN6_IFF_DEPRECATED is set from the
1016 	 * userland, make it deprecated.
1017 	 */
1018 	if ((ifra->ifra_flags & IN6_IFF_DEPRECATED) != 0) {
1019 		ia->ia6_lifetime.ia6t_pltime = 0;
1020 		ia->ia6_lifetime.ia6t_preferred = time_uptime;
1021 	}
1022 
1023 	/* reset the interface and routing table appropriately. */
1024 	if ((error = in6_ifinit(ifp, ia, &ifra->ifra_addr, hostIsNew)) != 0) {
1025 		if (hostIsNew)
1026 			free(ia, M_IFADDR);
1027 		goto exit;
1028 	}
1029 
1030 	/*
1031 	 * We are done if we have simply modified an existing address.
1032 	 */
1033 	if (!hostIsNew)
1034 		return error;
1035 
1036 	/*
1037 	 * Insert ia to the global list and ifa to the interface's list.
1038 	 */
1039 	IN6_ADDRLIST_WRITER_INSERT_TAIL(ia);
1040 
1041 	/* gain a refcnt for the link from in6_ifaddr */
1042 	ifaref(&ia->ia_ifa);
1043 	ifa_insert(ifp, &ia->ia_ifa);
1044 
1045 	/*
1046 	 * Beyond this point, we should call in6_purgeaddr upon an error,
1047 	 * not just go to unlink.
1048 	 */
1049 
1050 	/* join necessary multicast groups */
1051 	if ((ifp->if_flags & IFF_MULTICAST) != 0) {
1052 		struct sockaddr_in6 mltaddr, mltmask;
1053 		struct in6_addr llsol;
1054 
1055 		/* join solicited multicast addr for new host id */
1056 		memset(&llsol, 0, sizeof(struct in6_addr));
1057 		llsol.s6_addr16[0] = htons(0xff02);
1058 		llsol.s6_addr32[1] = 0;
1059 		llsol.s6_addr32[2] = htonl(1);
1060 		llsol.s6_addr32[3] = ifra->ifra_addr.sin6_addr.s6_addr32[3];
1061 		llsol.s6_addr8[12] = 0xff;
1062 		if ((error = in6_setscope(&llsol, ifp, NULL)) != 0) {
1063 			/* XXX: should not happen */
1064 			log(LOG_ERR, "%s: in6_setscope failed\n", __func__);
1065 			goto cleanup;
1066 		}
1067 		dad_delay = 0;
1068 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1069 			/*
1070 			 * We need a random delay for DAD on the address
1071 			 * being configured.  It also means delaying
1072 			 * transmission of the corresponding MLD report to
1073 			 * avoid report collision.
1074 			 * [draft-ietf-ipv6-rfc2462bis-02.txt]
1075 			 */
1076 			dad_delay = cprng_fast32() %
1077 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1078 		}
1079 
1080 #define	MLTMASK_LEN  4	/* mltmask's masklen (=32bit=4octet) */
1081 		/* join solicited multicast addr for new host id */
1082 		imm = in6_joingroup(ifp, &llsol, &error, dad_delay);
1083 		if (!imm) {
1084 			nd6log(LOG_ERR,
1085 			    "addmulti failed for %s on %s (errno=%d)\n",
1086 			    ip6_sprintf(&llsol), if_name(ifp), error);
1087 			goto cleanup;
1088 		}
1089 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1090 		in6m_sol = imm->i6mm_maddr;
1091 
1092 		sockaddr_in6_init(&mltmask, &in6mask32, 0, 0, 0);
1093 
1094 		/*
1095 		 * join link-local all-nodes address
1096 		 */
1097 		sockaddr_in6_init(&mltaddr, &in6addr_linklocal_allnodes,
1098 		    0, 0, 0);
1099 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1100 			goto cleanup; /* XXX: should not fail */
1101 
1102 		/*
1103 		 * XXX: do we really need this automatic routes?
1104 		 * We should probably reconsider this stuff.  Most applications
1105 		 * actually do not need the routes, since they usually specify
1106 		 * the outgoing interface.
1107 		 */
1108 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1109 		if (rt) {
1110 			if (memcmp(&mltaddr.sin6_addr,
1111 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1112 			    MLTMASK_LEN)) {
1113 				rtfree(rt);
1114 				rt = NULL;
1115 			} else if (rt->rt_ifp != ifp) {
1116 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1117 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1118 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1119 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1120 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1121 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1122 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1123 				rt_replace_ifa(rt, &ia->ia_ifa);
1124 				rt->rt_ifp = ifp;
1125 			}
1126 		}
1127 		if (!rt) {
1128 			struct rt_addrinfo info;
1129 
1130 			memset(&info, 0, sizeof(info));
1131 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1132 			info.rti_info[RTAX_GATEWAY] =
1133 			    (struct sockaddr *)&ia->ia_addr;
1134 			info.rti_info[RTAX_NETMASK] =
1135 			    (struct sockaddr *)&mltmask;
1136 			info.rti_info[RTAX_IFA] =
1137 			    (struct sockaddr *)&ia->ia_addr;
1138 			/* XXX: we need RTF_CONNECTED to fake nd6_rtrequest */
1139 			info.rti_flags = RTF_UP | RTF_CONNECTED;
1140 			error = rtrequest1(RTM_ADD, &info, NULL);
1141 			if (error)
1142 				goto cleanup;
1143 		} else {
1144 			rtfree(rt);
1145 		}
1146 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1147 		if (!imm) {
1148 			nd6log(LOG_WARNING,
1149 			    "addmulti failed for %s on %s (errno=%d)\n",
1150 			    ip6_sprintf(&mltaddr.sin6_addr),
1151 			    if_name(ifp), error);
1152 			goto cleanup;
1153 		}
1154 		LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1155 
1156 		/*
1157 		 * join node information group address
1158 		 */
1159 		dad_delay = 0;
1160 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1161 			/*
1162 			 * The spec doesn't say anything about delay for this
1163 			 * group, but the same logic should apply.
1164 			 */
1165 			dad_delay = cprng_fast32() %
1166 			    (MAX_RTR_SOLICITATION_DELAY * hz);
1167 		}
1168 		if (in6_nigroup(ifp, hostname, hostnamelen, &mltaddr) != 0)
1169 			;
1170 		else if ((imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error,
1171 		          dad_delay)) == NULL) { /* XXX jinmei */
1172 			nd6log(LOG_WARNING,
1173 			    "addmulti failed for %s on %s (errno=%d)\n",
1174 			    ip6_sprintf(&mltaddr.sin6_addr),
1175 			    if_name(ifp), error);
1176 			/* XXX not very fatal, go on... */
1177 		} else {
1178 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1179 		}
1180 
1181 
1182 		/*
1183 		 * join interface-local all-nodes address.
1184 		 * (ff01::1%ifN, and ff01::%ifN/32)
1185 		 */
1186 		mltaddr.sin6_addr = in6addr_nodelocal_allnodes;
1187 		if ((error = in6_setscope(&mltaddr.sin6_addr, ifp, NULL)) != 0)
1188 			goto cleanup; /* XXX: should not fail */
1189 
1190 		/* XXX: again, do we really need the route? */
1191 		rt = rtalloc1((struct sockaddr *)&mltaddr, 0);
1192 		if (rt) {
1193 			/* 32bit came from "mltmask" */
1194 			if (memcmp(&mltaddr.sin6_addr,
1195 			    &satocsin6(rt_getkey(rt))->sin6_addr,
1196 			    32 / NBBY)) {
1197 				rtfree(rt);
1198 				rt = NULL;
1199 			} else if (rt->rt_ifp != ifp) {
1200 				IN6_DPRINTF("%s: rt_ifp %p -> %p (%s) "
1201 				    "network %04x:%04x::/32 = %04x:%04x::/32\n",
1202 				    __func__, rt->rt_ifp, ifp, ifp->if_xname,
1203 				    ntohs(mltaddr.sin6_addr.s6_addr16[0]),
1204 				    ntohs(mltaddr.sin6_addr.s6_addr16[1]),
1205 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[0],
1206 				    satocsin6(rt_getkey(rt))->sin6_addr.s6_addr16[1]);
1207 				rt_replace_ifa(rt, &ia->ia_ifa);
1208 				rt->rt_ifp = ifp;
1209 			}
1210 		}
1211 		if (!rt) {
1212 			struct rt_addrinfo info;
1213 
1214 			memset(&info, 0, sizeof(info));
1215 			info.rti_info[RTAX_DST] = (struct sockaddr *)&mltaddr;
1216 			info.rti_info[RTAX_GATEWAY] =
1217 			    (struct sockaddr *)&ia->ia_addr;
1218 			info.rti_info[RTAX_NETMASK] =
1219 			    (struct sockaddr *)&mltmask;
1220 			info.rti_info[RTAX_IFA] =
1221 			    (struct sockaddr *)&ia->ia_addr;
1222 			info.rti_flags = RTF_UP | RTF_CONNECTED;
1223 			error = rtrequest1(RTM_ADD, &info, NULL);
1224 			if (error)
1225 				goto cleanup;
1226 #undef	MLTMASK_LEN
1227 		} else {
1228 			rtfree(rt);
1229 		}
1230 		imm = in6_joingroup(ifp, &mltaddr.sin6_addr, &error, 0);
1231 		if (!imm) {
1232 			nd6log(LOG_WARNING,
1233 			    "addmulti failed for %s on %s (errno=%d)\n",
1234 			    ip6_sprintf(&mltaddr.sin6_addr),
1235 			    if_name(ifp), error);
1236 			goto cleanup;
1237 		} else {
1238 			LIST_INSERT_HEAD(&ia->ia6_memberships, imm, i6mm_chain);
1239 		}
1240 	}
1241 
1242 	/* Add local address to lltable, if necessary (ex. on p2p link). */
1243 	error = nd6_add_ifa_lle(ia);
1244 	if (error != 0)
1245 		goto cleanup;
1246 
1247 	/*
1248 	 * Perform DAD, if needed.
1249 	 * XXX It may be of use, if we can administratively
1250 	 * disable DAD.
1251 	 */
1252 	if (hostIsNew && if_do_dad(ifp) &&
1253 	    ((ifra->ifra_flags & IN6_IFF_NODAD) == 0) &&
1254 	    (ia->ia6_flags & IN6_IFF_TENTATIVE))
1255 	{
1256 		int mindelay, maxdelay;
1257 
1258 		dad_delay = 0;
1259 		if ((flags & IN6_IFAUPDATE_DADDELAY)) {
1260 			/*
1261 			 * We need to impose a delay before sending an NS
1262 			 * for DAD.  Check if we also needed a delay for the
1263 			 * corresponding MLD message.  If we did, the delay
1264 			 * should be larger than the MLD delay (this could be
1265 			 * relaxed a bit, but this simple logic is at least
1266 			 * safe).
1267 			 */
1268 			mindelay = 0;
1269 			if (in6m_sol != NULL &&
1270 			    in6m_sol->in6m_state == MLD_REPORTPENDING) {
1271 				mindelay = in6m_sol->in6m_timer;
1272 			}
1273 			maxdelay = MAX_RTR_SOLICITATION_DELAY * hz;
1274 			if (maxdelay - mindelay == 0)
1275 				dad_delay = 0;
1276 			else {
1277 				dad_delay =
1278 				    (cprng_fast32() % (maxdelay - mindelay)) +
1279 				    mindelay;
1280 			}
1281 		}
1282 		/* +1 ensures callout is always used */
1283 		nd6_dad_start(&ia->ia_ifa, dad_delay + 1);
1284 	}
1285 
1286 	return 0;
1287 
1288   cleanup:
1289 	in6_purgeaddr(&ia->ia_ifa);
1290   exit:
1291 	return error;
1292 }
1293 
1294 int
1295 in6_update_ifa(struct ifnet *ifp, struct in6_aliasreq *ifra,
1296     struct in6_ifaddr *ia, int flags)
1297 {
1298 	int rc, s;
1299 
1300 	s = splnet();
1301 	rc = in6_update_ifa1(ifp, ifra, ia, flags);
1302 	splx(s);
1303 	return rc;
1304 }
1305 
1306 void
1307 in6_purgeaddr(struct ifaddr *ifa)
1308 {
1309 	struct ifnet *ifp = ifa->ifa_ifp;
1310 	struct in6_ifaddr *ia = (struct in6_ifaddr *) ifa;
1311 	struct in6_multi_mship *imm;
1312 
1313 	/* stop DAD processing */
1314 	nd6_dad_stop(ifa);
1315 
1316 	/*
1317 	 * delete route to the destination of the address being purged.
1318 	 * The interface must be p2p or loopback in this case.
1319 	 */
1320 	if ((ia->ia_flags & IFA_ROUTE) != 0 && ia->ia_dstaddr.sin6_len != 0) {
1321 		int e;
1322 
1323 		if ((e = rtinit(&(ia->ia_ifa), (int)RTM_DELETE, RTF_HOST))
1324 		    != 0) {
1325 			log(LOG_ERR, "in6_purgeaddr: failed to remove "
1326 			    "a route to the p2p destination: %s on %s, "
1327 			    "errno=%d\n",
1328 			    ip6_sprintf(&ia->ia_addr.sin6_addr), if_name(ifp),
1329 			    e);
1330 			/* proceed anyway... */
1331 		} else
1332 			ia->ia_flags &= ~IFA_ROUTE;
1333 	}
1334 
1335 	/* Remove ownaddr's loopback rtentry, if it exists. */
1336 	in6_ifremlocal(&(ia->ia_ifa));
1337 
1338 	/*
1339 	 * leave from multicast groups we have joined for the interface
1340 	 */
1341 	while ((imm = LIST_FIRST(&ia->ia6_memberships)) != NULL) {
1342 		LIST_REMOVE(imm, i6mm_chain);
1343 		in6_leavegroup(imm);
1344 	}
1345 
1346 	in6_unlink_ifa(ia, ifp);
1347 }
1348 
1349 static void
1350 in6_unlink_ifa(struct in6_ifaddr *ia, struct ifnet *ifp)
1351 {
1352 	int	s = splnet();
1353 
1354 	ifa_remove(ifp, &ia->ia_ifa);
1355 
1356 	IN6_ADDRLIST_WRITER_REMOVE(ia);
1357 	/* TODO psref_target_destroy */
1358 	IN6_ADDRLIST_ENTRY_DESTROY(ia);
1359 
1360 	/*
1361 	 * XXX thorpej@NetBSD.org -- if the interface is going
1362 	 * XXX away, don't save the multicast entries, delete them!
1363 	 */
1364 	if (LIST_EMPTY(&ia->ia6_multiaddrs))
1365 		;
1366 	else if (if_is_deactivated(ia->ia_ifa.ifa_ifp)) {
1367 		struct in6_multi *in6m, *next;
1368 
1369 		for (in6m = LIST_FIRST(&ia->ia6_multiaddrs); in6m != NULL;
1370 		     in6m = next) {
1371 			next = LIST_NEXT(in6m, in6m_entry);
1372 			in6_delmulti(in6m);
1373 		}
1374 	} else
1375 		in6_savemkludge(ia);
1376 
1377 	/*
1378 	 * Release the reference to the base prefix.  There should be a
1379 	 * positive reference.
1380 	 */
1381 	if (ia->ia6_ndpr == NULL) {
1382 		nd6log(LOG_NOTICE, "autoconf'ed address %p has no prefix\n",
1383 		    ia);
1384 	} else {
1385 		ia->ia6_ndpr->ndpr_refcnt--;
1386 		ia->ia6_ndpr = NULL;
1387 	}
1388 
1389 	/*
1390 	 * Also, if the address being removed is autoconf'ed, call
1391 	 * pfxlist_onlink_check() since the release might affect the status of
1392 	 * other (detached) addresses.
1393 	 */
1394 	if ((ia->ia6_flags & IN6_IFF_AUTOCONF) != 0)
1395 		pfxlist_onlink_check();
1396 
1397 	/*
1398 	 * release another refcnt for the link from in6_ifaddr.
1399 	 * Note that we should decrement the refcnt at least once for all *BSD.
1400 	 */
1401 	ifafree(&ia->ia_ifa);
1402 
1403 	splx(s);
1404 }
1405 
1406 void
1407 in6_purgeif(struct ifnet *ifp)
1408 {
1409 
1410 	in6_ifdetach(ifp);
1411 }
1412 
1413 /*
1414  * SIOC[GAD]LIFADDR.
1415  *	SIOCGLIFADDR: get first address. (?)
1416  *	SIOCGLIFADDR with IFLR_PREFIX:
1417  *		get first address that matches the specified prefix.
1418  *	SIOCALIFADDR: add the specified address.
1419  *	SIOCALIFADDR with IFLR_PREFIX:
1420  *		add the specified prefix, filling hostid part from
1421  *		the first link-local address.  prefixlen must be <= 64.
1422  *	SIOCDLIFADDR: delete the specified address.
1423  *	SIOCDLIFADDR with IFLR_PREFIX:
1424  *		delete the first address that matches the specified prefix.
1425  * return values:
1426  *	EINVAL on invalid parameters
1427  *	EADDRNOTAVAIL on prefix match failed/specified address not found
1428  *	other values may be returned from in6_ioctl()
1429  *
1430  * NOTE: SIOCALIFADDR(with IFLR_PREFIX set) allows prefixlen less than 64.
1431  * this is to accommodate address naming scheme other than RFC2374,
1432  * in the future.
1433  * RFC2373 defines interface id to be 64bit, but it allows non-RFC2374
1434  * address encoding scheme. (see figure on page 8)
1435  */
1436 static int
1437 in6_lifaddr_ioctl(struct socket *so, u_long cmd, void *data,
1438 	struct ifnet *ifp)
1439 {
1440 	struct in6_ifaddr *ia;
1441 	struct if_laddrreq *iflr = (struct if_laddrreq *)data;
1442 	struct ifaddr *ifa;
1443 	struct sockaddr *sa;
1444 
1445 	/* sanity checks */
1446 	if (!data || !ifp) {
1447 		panic("invalid argument to in6_lifaddr_ioctl");
1448 		/* NOTREACHED */
1449 	}
1450 
1451 	switch (cmd) {
1452 	case SIOCGLIFADDR:
1453 		/* address must be specified on GET with IFLR_PREFIX */
1454 		if ((iflr->flags & IFLR_PREFIX) == 0)
1455 			break;
1456 		/* FALLTHROUGH */
1457 	case SIOCALIFADDR:
1458 	case SIOCDLIFADDR:
1459 		/* address must be specified on ADD and DELETE */
1460 		sa = (struct sockaddr *)&iflr->addr;
1461 		if (sa->sa_family != AF_INET6)
1462 			return EINVAL;
1463 		if (sa->sa_len != sizeof(struct sockaddr_in6))
1464 			return EINVAL;
1465 		/* XXX need improvement */
1466 		sa = (struct sockaddr *)&iflr->dstaddr;
1467 		if (sa->sa_family && sa->sa_family != AF_INET6)
1468 			return EINVAL;
1469 		if (sa->sa_len && sa->sa_len != sizeof(struct sockaddr_in6))
1470 			return EINVAL;
1471 		break;
1472 	default: /* shouldn't happen */
1473 #if 0
1474 		panic("invalid cmd to in6_lifaddr_ioctl");
1475 		/* NOTREACHED */
1476 #else
1477 		return EOPNOTSUPP;
1478 #endif
1479 	}
1480 	if (sizeof(struct in6_addr) * NBBY < iflr->prefixlen)
1481 		return EINVAL;
1482 
1483 	switch (cmd) {
1484 	case SIOCALIFADDR:
1485 	    {
1486 		struct in6_aliasreq ifra;
1487 		struct in6_addr *xhostid = NULL;
1488 		int prefixlen;
1489 
1490 		if ((iflr->flags & IFLR_PREFIX) != 0) {
1491 			struct sockaddr_in6 *sin6;
1492 
1493 			/*
1494 			 * xhostid is to fill in the hostid part of the
1495 			 * address.  xhostid points to the first link-local
1496 			 * address attached to the interface.
1497 			 */
1498 			ia = in6ifa_ifpforlinklocal(ifp, 0);
1499 			if (ia == NULL)
1500 				return EADDRNOTAVAIL;
1501 			xhostid = IFA_IN6(&ia->ia_ifa);
1502 
1503 		 	/* prefixlen must be <= 64. */
1504 			if (64 < iflr->prefixlen)
1505 				return EINVAL;
1506 			prefixlen = iflr->prefixlen;
1507 
1508 			/* hostid part must be zero. */
1509 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1510 			if (sin6->sin6_addr.s6_addr32[2] != 0
1511 			 || sin6->sin6_addr.s6_addr32[3] != 0) {
1512 				return EINVAL;
1513 			}
1514 		} else
1515 			prefixlen = iflr->prefixlen;
1516 
1517 		/* copy args to in6_aliasreq, perform ioctl(SIOCAIFADDR_IN6). */
1518 		memset(&ifra, 0, sizeof(ifra));
1519 		memcpy(ifra.ifra_name, iflr->iflr_name, sizeof(ifra.ifra_name));
1520 
1521 		memcpy(&ifra.ifra_addr, &iflr->addr,
1522 		    ((struct sockaddr *)&iflr->addr)->sa_len);
1523 		if (xhostid) {
1524 			/* fill in hostid part */
1525 			ifra.ifra_addr.sin6_addr.s6_addr32[2] =
1526 			    xhostid->s6_addr32[2];
1527 			ifra.ifra_addr.sin6_addr.s6_addr32[3] =
1528 			    xhostid->s6_addr32[3];
1529 		}
1530 
1531 		if (((struct sockaddr *)&iflr->dstaddr)->sa_family) { /* XXX */
1532 			memcpy(&ifra.ifra_dstaddr, &iflr->dstaddr,
1533 			    ((struct sockaddr *)&iflr->dstaddr)->sa_len);
1534 			if (xhostid) {
1535 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[2] =
1536 				    xhostid->s6_addr32[2];
1537 				ifra.ifra_dstaddr.sin6_addr.s6_addr32[3] =
1538 				    xhostid->s6_addr32[3];
1539 			}
1540 		}
1541 
1542 		ifra.ifra_prefixmask.sin6_len = sizeof(struct sockaddr_in6);
1543 		in6_prefixlen2mask(&ifra.ifra_prefixmask.sin6_addr, prefixlen);
1544 
1545 		ifra.ifra_lifetime.ia6t_vltime = ND6_INFINITE_LIFETIME;
1546 		ifra.ifra_lifetime.ia6t_pltime = ND6_INFINITE_LIFETIME;
1547 		ifra.ifra_flags = iflr->flags & ~IFLR_PREFIX;
1548 		return in6_control(so, SIOCAIFADDR_IN6, &ifra, ifp);
1549 	    }
1550 	case SIOCGLIFADDR:
1551 	case SIOCDLIFADDR:
1552 	    {
1553 		struct in6_addr mask, candidate, match;
1554 		struct sockaddr_in6 *sin6;
1555 		int cmp;
1556 
1557 		memset(&mask, 0, sizeof(mask));
1558 		if (iflr->flags & IFLR_PREFIX) {
1559 			/* lookup a prefix rather than address. */
1560 			in6_prefixlen2mask(&mask, iflr->prefixlen);
1561 
1562 			sin6 = (struct sockaddr_in6 *)&iflr->addr;
1563 			memcpy(&match, &sin6->sin6_addr, sizeof(match));
1564 			match.s6_addr32[0] &= mask.s6_addr32[0];
1565 			match.s6_addr32[1] &= mask.s6_addr32[1];
1566 			match.s6_addr32[2] &= mask.s6_addr32[2];
1567 			match.s6_addr32[3] &= mask.s6_addr32[3];
1568 
1569 			/* if you set extra bits, that's wrong */
1570 			if (memcmp(&match, &sin6->sin6_addr, sizeof(match)))
1571 				return EINVAL;
1572 
1573 			cmp = 1;
1574 		} else {
1575 			if (cmd == SIOCGLIFADDR) {
1576 				/* on getting an address, take the 1st match */
1577 				cmp = 0;	/* XXX */
1578 			} else {
1579 				/* on deleting an address, do exact match */
1580 				in6_prefixlen2mask(&mask, 128);
1581 				sin6 = (struct sockaddr_in6 *)&iflr->addr;
1582 				memcpy(&match, &sin6->sin6_addr, sizeof(match));
1583 
1584 				cmp = 1;
1585 			}
1586 		}
1587 
1588 		IFADDR_READER_FOREACH(ifa, ifp) {
1589 			if (ifa->ifa_addr->sa_family != AF_INET6)
1590 				continue;
1591 			if (!cmp)
1592 				break;
1593 
1594 			/*
1595 			 * XXX: this is adhoc, but is necessary to allow
1596 			 * a user to specify fe80::/64 (not /10) for a
1597 			 * link-local address.
1598 			 */
1599 			memcpy(&candidate, IFA_IN6(ifa), sizeof(candidate));
1600 			in6_clearscope(&candidate);
1601 			candidate.s6_addr32[0] &= mask.s6_addr32[0];
1602 			candidate.s6_addr32[1] &= mask.s6_addr32[1];
1603 			candidate.s6_addr32[2] &= mask.s6_addr32[2];
1604 			candidate.s6_addr32[3] &= mask.s6_addr32[3];
1605 			if (IN6_ARE_ADDR_EQUAL(&candidate, &match))
1606 				break;
1607 		}
1608 		if (!ifa)
1609 			return EADDRNOTAVAIL;
1610 		ia = ifa2ia6(ifa);
1611 
1612 		if (cmd == SIOCGLIFADDR) {
1613 			int error;
1614 
1615 			/* fill in the if_laddrreq structure */
1616 			memcpy(&iflr->addr, &ia->ia_addr, ia->ia_addr.sin6_len);
1617 			error = sa6_recoverscope(
1618 			    (struct sockaddr_in6 *)&iflr->addr);
1619 			if (error != 0)
1620 				return error;
1621 
1622 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1623 				memcpy(&iflr->dstaddr, &ia->ia_dstaddr,
1624 				    ia->ia_dstaddr.sin6_len);
1625 				error = sa6_recoverscope(
1626 				    (struct sockaddr_in6 *)&iflr->dstaddr);
1627 				if (error != 0)
1628 					return error;
1629 			} else
1630 				memset(&iflr->dstaddr, 0, sizeof(iflr->dstaddr));
1631 
1632 			iflr->prefixlen =
1633 			    in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL);
1634 
1635 			iflr->flags = ia->ia6_flags;	/* XXX */
1636 
1637 			return 0;
1638 		} else {
1639 			struct in6_aliasreq ifra;
1640 
1641 			/* fill in6_aliasreq and do ioctl(SIOCDIFADDR_IN6) */
1642 			memset(&ifra, 0, sizeof(ifra));
1643 			memcpy(ifra.ifra_name, iflr->iflr_name,
1644 			    sizeof(ifra.ifra_name));
1645 
1646 			memcpy(&ifra.ifra_addr, &ia->ia_addr,
1647 			    ia->ia_addr.sin6_len);
1648 			if ((ifp->if_flags & IFF_POINTOPOINT) != 0) {
1649 				memcpy(&ifra.ifra_dstaddr, &ia->ia_dstaddr,
1650 				    ia->ia_dstaddr.sin6_len);
1651 			} else {
1652 				memset(&ifra.ifra_dstaddr, 0,
1653 				    sizeof(ifra.ifra_dstaddr));
1654 			}
1655 			memcpy(&ifra.ifra_dstaddr, &ia->ia_prefixmask,
1656 			    ia->ia_prefixmask.sin6_len);
1657 
1658 			ifra.ifra_flags = ia->ia6_flags;
1659 			return in6_control(so, SIOCDIFADDR_IN6, &ifra, ifp);
1660 		}
1661 	    }
1662 	}
1663 
1664 	return EOPNOTSUPP;	/* just for safety */
1665 }
1666 
1667 /*
1668  * Initialize an interface's internet6 address
1669  * and routing table entry.
1670  */
1671 static int
1672 in6_ifinit(struct ifnet *ifp, struct in6_ifaddr *ia,
1673 	const struct sockaddr_in6 *sin6, int newhost)
1674 {
1675 	int	error = 0, plen, ifacount = 0;
1676 	int	s = splnet();
1677 	struct ifaddr *ifa;
1678 
1679 	/*
1680 	 * Give the interface a chance to initialize
1681 	 * if this is its first address,
1682 	 * and to validate the address if necessary.
1683 	 */
1684 	IFADDR_READER_FOREACH(ifa, ifp) {
1685 		if (ifa->ifa_addr->sa_family != AF_INET6)
1686 			continue;
1687 		ifacount++;
1688 	}
1689 
1690 	ia->ia_addr = *sin6;
1691 
1692 	if (ifacount <= 0 &&
1693 	    (error = if_addr_init(ifp, &ia->ia_ifa, true)) != 0) {
1694 		splx(s);
1695 		return error;
1696 	}
1697 	splx(s);
1698 
1699 	ia->ia_ifa.ifa_metric = ifp->if_metric;
1700 
1701 	/* we could do in(6)_socktrim here, but just omit it at this moment. */
1702 
1703 	/*
1704 	 * Special case:
1705 	 * If the destination address is specified for a point-to-point
1706 	 * interface, install a route to the destination as an interface
1707 	 * direct route.
1708 	 */
1709 	plen = in6_mask2len(&ia->ia_prefixmask.sin6_addr, NULL); /* XXX */
1710 	if (plen == 128 && ia->ia_dstaddr.sin6_family == AF_INET6) {
1711 		if ((error = rtinit(&ia->ia_ifa, RTM_ADD,
1712 				    RTF_UP | RTF_HOST)) != 0)
1713 			return error;
1714 		ia->ia_flags |= IFA_ROUTE;
1715 	}
1716 
1717 	/* Add ownaddr as loopback rtentry, if necessary (ex. on p2p link). */
1718 	if (newhost) {
1719 		/* set the rtrequest function to create llinfo */
1720 		if (ifp->if_flags & IFF_POINTOPOINT)
1721 			ia->ia_ifa.ifa_rtrequest = p2p_rtrequest;
1722 		else if ((ifp->if_flags & IFF_LOOPBACK) == 0)
1723 			ia->ia_ifa.ifa_rtrequest = nd6_rtrequest;
1724 		in6_ifaddlocal(&ia->ia_ifa);
1725 	} else {
1726 		/* Inform the routing socket of new flags/timings */
1727 		rt_newaddrmsg(RTM_NEWADDR, &ia->ia_ifa, 0, NULL);
1728 	}
1729 
1730 	if (ifp->if_flags & IFF_MULTICAST)
1731 		in6_restoremkludge(ia, ifp);
1732 
1733 	return error;
1734 }
1735 
1736 static struct ifaddr *
1737 bestifa(struct ifaddr *best_ifa, struct ifaddr *ifa)
1738 {
1739 	if (best_ifa == NULL || best_ifa->ifa_preference < ifa->ifa_preference)
1740 		return ifa;
1741 	return best_ifa;
1742 }
1743 
1744 /*
1745  * Find an IPv6 interface link-local address specific to an interface.
1746  */
1747 struct in6_ifaddr *
1748 in6ifa_ifpforlinklocal(const struct ifnet *ifp, const int ignoreflags)
1749 {
1750 	struct ifaddr *best_ifa = NULL, *ifa;
1751 
1752 	IFADDR_READER_FOREACH(ifa, ifp) {
1753 		if (ifa->ifa_addr->sa_family != AF_INET6)
1754 			continue;
1755 		if (!IN6_IS_ADDR_LINKLOCAL(IFA_IN6(ifa)))
1756 			continue;
1757 		if ((((struct in6_ifaddr *)ifa)->ia6_flags & ignoreflags) != 0)
1758 			continue;
1759 		best_ifa = bestifa(best_ifa, ifa);
1760 	}
1761 
1762 	return (struct in6_ifaddr *)best_ifa;
1763 }
1764 
1765 /*
1766  * find the internet address corresponding to a given address.
1767  * ifaddr is returned referenced.
1768  */
1769 struct in6_ifaddr *
1770 in6ifa_ifwithaddr(const struct in6_addr *addr, uint32_t zoneid)
1771 {
1772 	struct in6_ifaddr *ia;
1773 
1774 #ifdef __FreeBSD__
1775 	IN6_IFADDR_RLOCK();
1776 	LIST_FOREACH(ia, IN6ADDR_HASH(addr), ia6_hash) {
1777 #else
1778 	IN6_ADDRLIST_READER_FOREACH(ia) {
1779 #endif
1780 		if (IN6_ARE_ADDR_EQUAL(IA6_IN6(ia), addr)) {
1781 			if (zoneid != 0 &&
1782 			    zoneid != ia->ia_addr.sin6_scope_id)
1783 				continue;
1784 			ifaref(&ia->ia_ifa);
1785 			break;
1786 		}
1787 	}
1788 #ifdef __FreeBSD__
1789 	IN6_IFADDR_RUNLOCK();
1790 #endif
1791 	return ia;
1792 }
1793 
1794 /*
1795  * find the internet address corresponding to a given interface and address.
1796  */
1797 struct in6_ifaddr *
1798 in6ifa_ifpwithaddr(const struct ifnet *ifp, const struct in6_addr *addr)
1799 {
1800 	struct ifaddr *best_ifa = NULL, *ifa;
1801 
1802 	IFADDR_READER_FOREACH(ifa, ifp) {
1803 		if (ifa->ifa_addr->sa_family != AF_INET6)
1804 			continue;
1805 		if (!IN6_ARE_ADDR_EQUAL(addr, IFA_IN6(ifa)))
1806 			continue;
1807 		best_ifa = bestifa(best_ifa, ifa);
1808 	}
1809 
1810 	return (struct in6_ifaddr *)best_ifa;
1811 }
1812 
1813 static struct in6_ifaddr *
1814 bestia(struct in6_ifaddr *best_ia, struct in6_ifaddr *ia)
1815 {
1816 	if (best_ia == NULL ||
1817 	    best_ia->ia_ifa.ifa_preference < ia->ia_ifa.ifa_preference)
1818 		return ia;
1819 	return best_ia;
1820 }
1821 
1822 /*
1823  * Convert IP6 address to printable (loggable) representation.
1824  */
1825 char *
1826 ip6_sprintf(const struct in6_addr *addr)
1827 {
1828 	static int ip6round = 0;
1829 	static char ip6buf[8][INET6_ADDRSTRLEN];
1830 	char *cp = ip6buf[ip6round++ & 7];
1831 
1832 	in6_print(cp, INET6_ADDRSTRLEN, addr);
1833 	return cp;
1834 }
1835 
1836 /*
1837  * Determine if an address is on a local network.
1838  */
1839 int
1840 in6_localaddr(const struct in6_addr *in6)
1841 {
1842 	struct in6_ifaddr *ia;
1843 
1844 	if (IN6_IS_ADDR_LOOPBACK(in6) || IN6_IS_ADDR_LINKLOCAL(in6))
1845 		return 1;
1846 
1847 	IN6_ADDRLIST_READER_FOREACH(ia) {
1848 		if (IN6_ARE_MASKED_ADDR_EQUAL(in6, &ia->ia_addr.sin6_addr,
1849 					      &ia->ia_prefixmask.sin6_addr))
1850 			return 1;
1851 	}
1852 
1853 	return 0;
1854 }
1855 
1856 int
1857 in6_is_addr_deprecated(struct sockaddr_in6 *sa6)
1858 {
1859 	struct in6_ifaddr *ia;
1860 
1861 	IN6_ADDRLIST_READER_FOREACH(ia) {
1862 		if (IN6_ARE_ADDR_EQUAL(&ia->ia_addr.sin6_addr,
1863 		    &sa6->sin6_addr) &&
1864 #ifdef SCOPEDROUTING
1865 		    ia->ia_addr.sin6_scope_id == sa6->sin6_scope_id &&
1866 #endif
1867 		    (ia->ia6_flags & IN6_IFF_DEPRECATED) != 0)
1868 			return 1; /* true */
1869 
1870 		/* XXX: do we still have to go thru the rest of the list? */
1871 	}
1872 
1873 	return 0;		/* false */
1874 }
1875 
1876 /*
1877  * return length of part which dst and src are equal
1878  * hard coding...
1879  */
1880 int
1881 in6_matchlen(struct in6_addr *src, struct in6_addr *dst)
1882 {
1883 	int match = 0;
1884 	u_char *s = (u_char *)src, *d = (u_char *)dst;
1885 	u_char *lim = s + 16, r;
1886 
1887 	while (s < lim)
1888 		if ((r = (*d++ ^ *s++)) != 0) {
1889 			while (r < 128) {
1890 				match++;
1891 				r <<= 1;
1892 			}
1893 			break;
1894 		} else
1895 			match += NBBY;
1896 	return match;
1897 }
1898 
1899 /* XXX: to be scope conscious */
1900 int
1901 in6_are_prefix_equal(struct in6_addr *p1, struct in6_addr *p2, int len)
1902 {
1903 	int bytelen, bitlen;
1904 
1905 	/* sanity check */
1906 	if (len < 0 || len > 128) {
1907 		log(LOG_ERR, "in6_are_prefix_equal: invalid prefix length(%d)\n",
1908 		    len);
1909 		return 0;
1910 	}
1911 
1912 	bytelen = len / NBBY;
1913 	bitlen = len % NBBY;
1914 
1915 	if (memcmp(&p1->s6_addr, &p2->s6_addr, bytelen))
1916 		return 0;
1917 	if (bitlen != 0 &&
1918 	    p1->s6_addr[bytelen] >> (NBBY - bitlen) !=
1919 	    p2->s6_addr[bytelen] >> (NBBY - bitlen))
1920 		return 0;
1921 
1922 	return 1;
1923 }
1924 
1925 void
1926 in6_prefixlen2mask(struct in6_addr *maskp, int len)
1927 {
1928 	static const u_char maskarray[NBBY] = {0x80, 0xc0, 0xe0, 0xf0, 0xf8, 0xfc, 0xfe, 0xff};
1929 	int bytelen, bitlen, i;
1930 
1931 	/* sanity check */
1932 	if (len < 0 || len > 128) {
1933 		log(LOG_ERR, "in6_prefixlen2mask: invalid prefix length(%d)\n",
1934 		    len);
1935 		return;
1936 	}
1937 
1938 	memset(maskp, 0, sizeof(*maskp));
1939 	bytelen = len / NBBY;
1940 	bitlen = len % NBBY;
1941 	for (i = 0; i < bytelen; i++)
1942 		maskp->s6_addr[i] = 0xff;
1943 	if (bitlen)
1944 		maskp->s6_addr[bytelen] = maskarray[bitlen - 1];
1945 }
1946 
1947 /*
1948  * return the best address out of the same scope. if no address was
1949  * found, return the first valid address from designated IF.
1950  */
1951 struct in6_ifaddr *
1952 in6_ifawithifp(struct ifnet *ifp, struct in6_addr *dst)
1953 {
1954 	int dst_scope =	in6_addrscope(dst), blen = -1, tlen;
1955 	struct ifaddr *ifa;
1956 	struct in6_ifaddr *best_ia = NULL, *ia;
1957 	struct in6_ifaddr *dep[2];	/* last-resort: deprecated */
1958 
1959 	dep[0] = dep[1] = NULL;
1960 
1961 	/*
1962 	 * We first look for addresses in the same scope.
1963 	 * If there is one, return it.
1964 	 * If two or more, return one which matches the dst longest.
1965 	 * If none, return one of global addresses assigned other ifs.
1966 	 */
1967 	IFADDR_READER_FOREACH(ifa, ifp) {
1968 		if (ifa->ifa_addr->sa_family != AF_INET6)
1969 			continue;
1970 		ia = (struct in6_ifaddr *)ifa;
1971 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
1972 			continue; /* XXX: is there any case to allow anycast? */
1973 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
1974 			continue; /* don't use this interface */
1975 		if (ia->ia6_flags & IN6_IFF_DETACHED)
1976 			continue;
1977 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
1978 			if (ip6_use_deprecated)
1979 				dep[0] = ia;
1980 			continue;
1981 		}
1982 
1983 		if (dst_scope != in6_addrscope(IFA_IN6(ifa)))
1984 			continue;
1985 		/*
1986 		 * call in6_matchlen() as few as possible
1987 		 */
1988 		if (best_ia == NULL) {
1989 			best_ia = ia;
1990 			continue;
1991 		}
1992 		if (blen == -1)
1993 			blen = in6_matchlen(&best_ia->ia_addr.sin6_addr, dst);
1994 		tlen = in6_matchlen(IFA_IN6(ifa), dst);
1995 		if (tlen > blen) {
1996 			blen = tlen;
1997 			best_ia = ia;
1998 		} else if (tlen == blen)
1999 			best_ia = bestia(best_ia, ia);
2000 	}
2001 	if (best_ia != NULL)
2002 		return best_ia;
2003 
2004 	IFADDR_READER_FOREACH(ifa, ifp) {
2005 		if (ifa->ifa_addr->sa_family != AF_INET6)
2006 			continue;
2007 		ia = (struct in6_ifaddr *)ifa;
2008 		if (ia->ia6_flags & IN6_IFF_ANYCAST)
2009 			continue; /* XXX: is there any case to allow anycast? */
2010 		if (ia->ia6_flags & IN6_IFF_NOTREADY)
2011 			continue; /* don't use this interface */
2012 		if (ia->ia6_flags & IN6_IFF_DETACHED)
2013 			continue;
2014 		if (ia->ia6_flags & IN6_IFF_DEPRECATED) {
2015 			if (ip6_use_deprecated)
2016 				dep[1] = (struct in6_ifaddr *)ifa;
2017 			continue;
2018 		}
2019 
2020 		best_ia = bestia(best_ia, ia);
2021 	}
2022 	if (best_ia != NULL)
2023 		return best_ia;
2024 
2025 	/* use the last-resort values, that are, deprecated addresses */
2026 	if (dep[0])
2027 		return dep[0];
2028 	if (dep[1])
2029 		return dep[1];
2030 
2031 	return NULL;
2032 }
2033 
2034 /*
2035  * perform DAD when interface becomes IFF_UP.
2036  */
2037 void
2038 in6_if_link_up(struct ifnet *ifp)
2039 {
2040 	struct ifaddr *ifa;
2041 	struct in6_ifaddr *ia;
2042 
2043 	/* Ensure it's sane to run DAD */
2044 	if (ifp->if_link_state == LINK_STATE_DOWN)
2045 		return;
2046 	if ((ifp->if_flags & (IFF_UP|IFF_RUNNING)) != (IFF_UP|IFF_RUNNING))
2047 		return;
2048 
2049 	IFADDR_READER_FOREACH(ifa, ifp) {
2050 		if (ifa->ifa_addr->sa_family != AF_INET6)
2051 			continue;
2052 		ia = (struct in6_ifaddr *)ifa;
2053 
2054 		/* If detached then mark as tentative */
2055 		if (ia->ia6_flags & IN6_IFF_DETACHED) {
2056 			ia->ia6_flags &= ~IN6_IFF_DETACHED;
2057 			if (if_do_dad(ifp)) {
2058 				ia->ia6_flags |= IN6_IFF_TENTATIVE;
2059 				nd6log(LOG_ERR, "%s marked tentative\n",
2060 				    ip6_sprintf(&ia->ia_addr.sin6_addr));
2061 			} else if ((ia->ia6_flags & IN6_IFF_TENTATIVE) == 0)
2062 				rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2063 		}
2064 
2065 		if (ia->ia6_flags & IN6_IFF_TENTATIVE) {
2066 			int rand_delay;
2067 
2068 			/* Clear the duplicated flag as we're starting DAD. */
2069 			ia->ia6_flags &= ~IN6_IFF_DUPLICATED;
2070 
2071 			/*
2072 			 * The TENTATIVE flag was likely set by hand
2073 			 * beforehand, implicitly indicating the need for DAD.
2074 			 * We may be able to skip the random delay in this
2075 			 * case, but we impose delays just in case.
2076 			 */
2077 			rand_delay = cprng_fast32() %
2078 			    (MAX_RTR_SOLICITATION_DELAY * hz);
2079 			/* +1 ensures callout is always used */
2080 			nd6_dad_start(ifa, rand_delay + 1);
2081 		}
2082 	}
2083 
2084 	/* Restore any detached prefixes */
2085 	pfxlist_onlink_check();
2086 }
2087 
2088 void
2089 in6_if_up(struct ifnet *ifp)
2090 {
2091 
2092 	/*
2093 	 * special cases, like 6to4, are handled in in6_ifattach
2094 	 */
2095 	in6_ifattach(ifp, NULL);
2096 
2097 	/* interface may not support link state, so bring it up also */
2098 	in6_if_link_up(ifp);
2099 }
2100 
2101 /*
2102  * Mark all addresses as detached.
2103  */
2104 void
2105 in6_if_link_down(struct ifnet *ifp)
2106 {
2107 	struct ifaddr *ifa;
2108 	struct in6_ifaddr *ia;
2109 
2110 	/* Any prefixes on this interface should be detached as well */
2111 	pfxlist_onlink_check();
2112 
2113 	IFADDR_READER_FOREACH(ifa, ifp) {
2114 		if (ifa->ifa_addr->sa_family != AF_INET6)
2115 			continue;
2116 		ia = (struct in6_ifaddr *)ifa;
2117 
2118 		/* Stop DAD processing */
2119 		nd6_dad_stop(ifa);
2120 
2121 		/*
2122 		 * Mark the address as detached.
2123 		 * This satisfies RFC4862 Section 5.3, but we should apply
2124 		 * this logic to all addresses to be a good citizen and
2125 		 * avoid potential duplicated addresses.
2126 		 * When the interface comes up again, detached addresses
2127 		 * are marked tentative and DAD commences.
2128 		 */
2129 		if (!(ia->ia6_flags & IN6_IFF_DETACHED)) {
2130 			nd6log(LOG_DEBUG, "%s marked detached\n",
2131 			    ip6_sprintf(&ia->ia_addr.sin6_addr));
2132 			ia->ia6_flags |= IN6_IFF_DETACHED;
2133 			ia->ia6_flags &=
2134 			    ~(IN6_IFF_TENTATIVE | IN6_IFF_DUPLICATED);
2135 			rt_newaddrmsg(RTM_NEWADDR, ifa, 0, NULL);
2136 		}
2137 	}
2138 }
2139 
2140 void
2141 in6_if_down(struct ifnet *ifp)
2142 {
2143 
2144 	in6_if_link_down(ifp);
2145 }
2146 
2147 void
2148 in6_if_link_state_change(struct ifnet *ifp, int link_state)
2149 {
2150 
2151 	switch (link_state) {
2152 	case LINK_STATE_DOWN:
2153 		in6_if_link_down(ifp);
2154 		break;
2155 	case LINK_STATE_UP:
2156 		in6_if_link_up(ifp);
2157 		break;
2158 	}
2159 }
2160 
2161 /*
2162  * Calculate max IPv6 MTU through all the interfaces and store it
2163  * to in6_maxmtu.
2164  */
2165 void
2166 in6_setmaxmtu(void)
2167 {
2168 	unsigned long maxmtu = 0;
2169 	struct ifnet *ifp;
2170 	int s;
2171 
2172 	s = pserialize_read_enter();
2173 	IFNET_READER_FOREACH(ifp) {
2174 		/* this function can be called during ifnet initialization */
2175 		if (!ifp->if_afdata[AF_INET6])
2176 			continue;
2177 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
2178 		    IN6_LINKMTU(ifp) > maxmtu)
2179 			maxmtu = IN6_LINKMTU(ifp);
2180 	}
2181 	pserialize_read_exit(s);
2182 	if (maxmtu)	     /* update only when maxmtu is positive */
2183 		in6_maxmtu = maxmtu;
2184 }
2185 
2186 /*
2187  * Provide the length of interface identifiers to be used for the link attached
2188  * to the given interface.  The length should be defined in "IPv6 over
2189  * xxx-link" document.  Note that address architecture might also define
2190  * the length for a particular set of address prefixes, regardless of the
2191  * link type.  As clarified in rfc2462bis, those two definitions should be
2192  * consistent, and those really are as of August 2004.
2193  */
2194 int
2195 in6_if2idlen(struct ifnet *ifp)
2196 {
2197 	switch (ifp->if_type) {
2198 	case IFT_ETHER:		/* RFC2464 */
2199 	case IFT_PROPVIRTUAL:	/* XXX: no RFC. treat it as ether */
2200 	case IFT_L2VLAN:	/* ditto */
2201 	case IFT_IEEE80211:	/* ditto */
2202 	case IFT_FDDI:		/* RFC2467 */
2203 	case IFT_ISO88025:	/* RFC2470 (IPv6 over Token Ring) */
2204 	case IFT_PPP:		/* RFC2472 */
2205 	case IFT_ARCNET:	/* RFC2497 */
2206 	case IFT_FRELAY:	/* RFC2590 */
2207 	case IFT_IEEE1394:	/* RFC3146 */
2208 	case IFT_GIF:		/* draft-ietf-v6ops-mech-v2-07 */
2209 	case IFT_LOOP:		/* XXX: is this really correct? */
2210 		return 64;
2211 	default:
2212 		/*
2213 		 * Unknown link type:
2214 		 * It might be controversial to use the today's common constant
2215 		 * of 64 for these cases unconditionally.  For full compliance,
2216 		 * we should return an error in this case.  On the other hand,
2217 		 * if we simply miss the standard for the link type or a new
2218 		 * standard is defined for a new link type, the IFID length
2219 		 * is very likely to be the common constant.  As a compromise,
2220 		 * we always use the constant, but make an explicit notice
2221 		 * indicating the "unknown" case.
2222 		 */
2223 		printf("in6_if2idlen: unknown link type (%d)\n", ifp->if_type);
2224 		return 64;
2225 	}
2226 }
2227 
2228 struct in6_llentry {
2229 	struct llentry		base;
2230 };
2231 
2232 #define	IN6_LLTBL_DEFAULT_HSIZE	32
2233 #define	IN6_LLTBL_HASH(k, h) \
2234 	(((((((k >> 8) ^ k) >> 8) ^ k) >> 8) ^ k) & ((h) - 1))
2235 
2236 /*
2237  * Do actual deallocation of @lle.
2238  * Called by LLE_FREE_LOCKED when number of references
2239  * drops to zero.
2240  */
2241 static void
2242 in6_lltable_destroy_lle(struct llentry *lle)
2243 {
2244 
2245 	LLE_WUNLOCK(lle);
2246 	LLE_LOCK_DESTROY(lle);
2247 	kmem_intr_free(lle, sizeof(struct in6_llentry));
2248 }
2249 
2250 static struct llentry *
2251 in6_lltable_new(const struct in6_addr *addr6, u_int flags)
2252 {
2253 	struct in6_llentry *lle;
2254 
2255 	lle = kmem_intr_zalloc(sizeof(struct in6_llentry), KM_NOSLEEP);
2256 	if (lle == NULL)		/* NB: caller generates msg */
2257 		return NULL;
2258 
2259 	lle->base.r_l3addr.addr6 = *addr6;
2260 	lle->base.lle_refcnt = 1;
2261 	lle->base.lle_free = in6_lltable_destroy_lle;
2262 	LLE_LOCK_INIT(&lle->base);
2263 	callout_init(&lle->base.lle_timer, CALLOUT_MPSAFE);
2264 
2265 	return &lle->base;
2266 }
2267 
2268 static int
2269 in6_lltable_match_prefix(const struct sockaddr *prefix,
2270     const struct sockaddr *mask, u_int flags, struct llentry *lle)
2271 {
2272 	const struct sockaddr_in6 *pfx = (const struct sockaddr_in6 *)prefix;
2273 	const struct sockaddr_in6 *msk = (const struct sockaddr_in6 *)mask;
2274 
2275 	if (IN6_ARE_MASKED_ADDR_EQUAL(&lle->r_l3addr.addr6,
2276 	    &pfx->sin6_addr, &msk->sin6_addr) &&
2277 	    ((flags & LLE_STATIC) || !(lle->la_flags & LLE_STATIC)))
2278 		return 1;
2279 
2280 	return 0;
2281 }
2282 
2283 static void
2284 in6_lltable_free_entry(struct lltable *llt, struct llentry *lle)
2285 {
2286 	struct ifnet *ifp __diagused;
2287 
2288 	LLE_WLOCK_ASSERT(lle);
2289 	KASSERT(llt != NULL);
2290 
2291 	/* Unlink entry from table */
2292 	if ((lle->la_flags & LLE_LINKED) != 0) {
2293 
2294 		ifp = llt->llt_ifp;
2295 		IF_AFDATA_WLOCK_ASSERT(ifp);
2296 		lltable_unlink_entry(llt, lle);
2297 	}
2298 
2299 	KASSERT(mutex_owned(softnet_lock));
2300 	callout_halt(&lle->lle_timer, softnet_lock);
2301 	LLE_REMREF(lle);
2302 
2303 	llentry_free(lle);
2304 }
2305 
2306 static int
2307 in6_lltable_rtcheck(struct ifnet *ifp,
2308 		    u_int flags,
2309 		    const struct sockaddr *l3addr)
2310 {
2311 	struct rtentry *rt;
2312 
2313 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2314 	    "sin_family %d", l3addr->sa_family);
2315 
2316 	rt = rtalloc1(l3addr, 0);
2317 	if (rt == NULL || (rt->rt_flags & RTF_GATEWAY) || rt->rt_ifp != ifp) {
2318 		struct ifaddr *ifa;
2319 		/*
2320 		 * Create an ND6 cache for an IPv6 neighbor
2321 		 * that is not covered by our own prefix.
2322 		 */
2323 		/* XXX ifaof_ifpforaddr should take a const param */
2324 		ifa = ifaof_ifpforaddr(l3addr, ifp);
2325 		if (ifa != NULL) {
2326 			ifafree(ifa);
2327 			if (rt != NULL)
2328 				rtfree(rt);
2329 			return 0;
2330 		}
2331 		log(LOG_INFO, "IPv6 address: \"%s\" is not on the network\n",
2332 		    ip6_sprintf(&((const struct sockaddr_in6 *)l3addr)->sin6_addr));
2333 		if (rt != NULL)
2334 			rtfree(rt);
2335 		return EINVAL;
2336 	}
2337 	rtfree(rt);
2338 	return 0;
2339 }
2340 
2341 static inline uint32_t
2342 in6_lltable_hash_dst(const struct in6_addr *dst, uint32_t hsize)
2343 {
2344 
2345 	return IN6_LLTBL_HASH(dst->s6_addr32[3], hsize);
2346 }
2347 
2348 static uint32_t
2349 in6_lltable_hash(const struct llentry *lle, uint32_t hsize)
2350 {
2351 
2352 	return in6_lltable_hash_dst(&lle->r_l3addr.addr6, hsize);
2353 }
2354 
2355 static void
2356 in6_lltable_fill_sa_entry(const struct llentry *lle, struct sockaddr *sa)
2357 {
2358 	struct sockaddr_in6 *sin6;
2359 
2360 	sin6 = (struct sockaddr_in6 *)sa;
2361 	bzero(sin6, sizeof(*sin6));
2362 	sin6->sin6_family = AF_INET6;
2363 	sin6->sin6_len = sizeof(*sin6);
2364 	sin6->sin6_addr = lle->r_l3addr.addr6;
2365 }
2366 
2367 static inline struct llentry *
2368 in6_lltable_find_dst(struct lltable *llt, const struct in6_addr *dst)
2369 {
2370 	struct llentry *lle;
2371 	struct llentries *lleh;
2372 	u_int hashidx;
2373 
2374 	hashidx = in6_lltable_hash_dst(dst, llt->llt_hsize);
2375 	lleh = &llt->lle_head[hashidx];
2376 	LIST_FOREACH(lle, lleh, lle_next) {
2377 		if (lle->la_flags & LLE_DELETED)
2378 			continue;
2379 		if (IN6_ARE_ADDR_EQUAL(&lle->r_l3addr.addr6, dst))
2380 			break;
2381 	}
2382 
2383 	return lle;
2384 }
2385 
2386 static int
2387 in6_lltable_delete(struct lltable *llt, u_int flags,
2388 	const struct sockaddr *l3addr)
2389 {
2390 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2391 	struct llentry *lle;
2392 
2393 	IF_AFDATA_WLOCK_ASSERT(llt->llt_ifp);
2394 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2395 	    "sin_family %d", l3addr->sa_family);
2396 
2397 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2398 
2399 	if (lle == NULL)
2400 		return ENOENT;
2401 
2402 	LLE_WLOCK(lle);
2403 	lle->la_flags |= LLE_DELETED;
2404 #ifdef DIAGNOSTIC
2405 	log(LOG_INFO, "ifaddr cache = %p is deleted\n", lle);
2406 #endif
2407 	if ((lle->la_flags & (LLE_STATIC | LLE_IFADDR)) == LLE_STATIC)
2408 		llentry_free(lle);
2409 	else
2410 		LLE_WUNLOCK(lle);
2411 
2412 	return 0;
2413 }
2414 
2415 static struct llentry *
2416 in6_lltable_create(struct lltable *llt, u_int flags,
2417 	const struct sockaddr *l3addr)
2418 {
2419 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2420 	struct ifnet *ifp = llt->llt_ifp;
2421 	struct llentry *lle;
2422 
2423 	IF_AFDATA_WLOCK_ASSERT(ifp);
2424 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2425 	    "sin_family %d", l3addr->sa_family);
2426 
2427 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2428 
2429 	if (lle != NULL) {
2430 		LLE_WLOCK(lle);
2431 		return lle;
2432 	}
2433 
2434 	/*
2435 	 * A route that covers the given address must have
2436 	 * been installed 1st because we are doing a resolution,
2437 	 * verify this.
2438 	 */
2439 	if (!(flags & LLE_IFADDR) &&
2440 	    in6_lltable_rtcheck(ifp, flags, l3addr) != 0)
2441 		return NULL;
2442 
2443 	lle = in6_lltable_new(&sin6->sin6_addr, flags);
2444 	if (lle == NULL) {
2445 		log(LOG_INFO, "lla_lookup: new lle malloc failed\n");
2446 		return NULL;
2447 	}
2448 	lle->la_flags = flags;
2449 	if ((flags & LLE_IFADDR) == LLE_IFADDR) {
2450 		memcpy(&lle->ll_addr, CLLADDR(ifp->if_sadl), ifp->if_addrlen);
2451 		lle->la_flags |= LLE_VALID;
2452 	}
2453 
2454 	lltable_link_entry(llt, lle);
2455 	LLE_WLOCK(lle);
2456 
2457 	return lle;
2458 }
2459 
2460 static struct llentry *
2461 in6_lltable_lookup(struct lltable *llt, u_int flags,
2462 	const struct sockaddr *l3addr)
2463 {
2464 	const struct sockaddr_in6 *sin6 = (const struct sockaddr_in6 *)l3addr;
2465 	struct llentry *lle;
2466 
2467 	IF_AFDATA_LOCK_ASSERT(llt->llt_ifp);
2468 	KASSERTMSG(l3addr->sa_family == AF_INET6,
2469 	    "sin_family %d", l3addr->sa_family);
2470 
2471 	lle = in6_lltable_find_dst(llt, &sin6->sin6_addr);
2472 
2473 	if (lle == NULL)
2474 		return NULL;
2475 
2476 	if (flags & LLE_EXCLUSIVE)
2477 		LLE_WLOCK(lle);
2478 	else
2479 		LLE_RLOCK(lle);
2480 	return lle;
2481 }
2482 
2483 static int
2484 in6_lltable_dump_entry(struct lltable *llt, struct llentry *lle,
2485     struct rt_walkarg *w)
2486 {
2487 	struct sockaddr_in6 sin6;
2488 
2489 	LLTABLE_LOCK_ASSERT();
2490 
2491 	/* skip deleted entries */
2492 	if (lle->la_flags & LLE_DELETED)
2493 		return 0;
2494 
2495 	sockaddr_in6_init(&sin6, &lle->r_l3addr.addr6, 0, 0, 0);
2496 
2497 	return lltable_dump_entry(llt, lle, w, sin6tosa(&sin6));
2498 }
2499 
2500 static struct lltable *
2501 in6_lltattach(struct ifnet *ifp)
2502 {
2503 	struct lltable *llt;
2504 
2505 	llt = lltable_allocate_htbl(IN6_LLTBL_DEFAULT_HSIZE);
2506 	llt->llt_af = AF_INET6;
2507 	llt->llt_ifp = ifp;
2508 
2509 	llt->llt_lookup = in6_lltable_lookup;
2510 	llt->llt_create = in6_lltable_create;
2511 	llt->llt_delete = in6_lltable_delete;
2512 	llt->llt_dump_entry = in6_lltable_dump_entry;
2513 	llt->llt_hash = in6_lltable_hash;
2514 	llt->llt_fill_sa_entry = in6_lltable_fill_sa_entry;
2515 	llt->llt_free_entry = in6_lltable_free_entry;
2516 	llt->llt_match_prefix = in6_lltable_match_prefix;
2517 	lltable_link(llt);
2518 
2519 	return llt;
2520 }
2521 
2522 void *
2523 in6_domifattach(struct ifnet *ifp)
2524 {
2525 	struct in6_ifextra *ext;
2526 
2527 	ext = malloc(sizeof(*ext), M_IFADDR, M_WAITOK|M_ZERO);
2528 
2529 	ext->in6_ifstat = malloc(sizeof(struct in6_ifstat),
2530 	    M_IFADDR, M_WAITOK|M_ZERO);
2531 
2532 	ext->icmp6_ifstat = malloc(sizeof(struct icmp6_ifstat),
2533 	    M_IFADDR, M_WAITOK|M_ZERO);
2534 
2535 	ext->nd_ifinfo = nd6_ifattach(ifp);
2536 	ext->scope6_id = scope6_ifattach(ifp);
2537 	ext->nprefixes = 0;
2538 	ext->ndefrouters = 0;
2539 
2540 	ext->lltable = in6_lltattach(ifp);
2541 
2542 	return ext;
2543 }
2544 
2545 void
2546 in6_domifdetach(struct ifnet *ifp, void *aux)
2547 {
2548 	struct in6_ifextra *ext = (struct in6_ifextra *)aux;
2549 
2550 	lltable_free(ext->lltable);
2551 	ext->lltable = NULL;
2552 	nd6_ifdetach(ifp, ext);
2553 	free(ext->in6_ifstat, M_IFADDR);
2554 	free(ext->icmp6_ifstat, M_IFADDR);
2555 	scope6_ifdetach(ext->scope6_id);
2556 	free(ext, M_IFADDR);
2557 }
2558 
2559 /*
2560  * Convert IPv4 address stored in struct in_addr to IPv4-Mapped IPv6 address
2561  * stored in struct in6_addr as defined in RFC 4921 section 2.5.5.2.
2562  */
2563 void
2564 in6_in_2_v4mapin6(const struct in_addr *in, struct in6_addr *in6)
2565 {
2566 	in6->s6_addr32[0] = 0;
2567 	in6->s6_addr32[1] = 0;
2568 	in6->s6_addr32[2] = IPV6_ADDR_INT32_SMP;
2569 	in6->s6_addr32[3] = in->s_addr;
2570 }
2571 
2572 /*
2573  * Convert sockaddr_in6 to sockaddr_in.  Original sockaddr_in6 must be
2574  * v4 mapped addr or v4 compat addr
2575  */
2576 void
2577 in6_sin6_2_sin(struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2578 {
2579 	memset(sin, 0, sizeof(*sin));
2580 	sin->sin_len = sizeof(struct sockaddr_in);
2581 	sin->sin_family = AF_INET;
2582 	sin->sin_port = sin6->sin6_port;
2583 	sin->sin_addr.s_addr = sin6->sin6_addr.s6_addr32[3];
2584 }
2585 
2586 /* Convert sockaddr_in to sockaddr_in6 in v4 mapped addr format. */
2587 void
2588 in6_sin_2_v4mapsin6(const struct sockaddr_in *sin, struct sockaddr_in6 *sin6)
2589 {
2590 	memset(sin6, 0, sizeof(*sin6));
2591 	sin6->sin6_len = sizeof(struct sockaddr_in6);
2592 	sin6->sin6_family = AF_INET6;
2593 	sin6->sin6_port = sin->sin_port;
2594 	in6_in_2_v4mapin6(&sin->sin_addr, &sin6->sin6_addr);
2595 }
2596 
2597 /* Convert sockaddr_in6 into sockaddr_in. */
2598 void
2599 in6_sin6_2_sin_in_sock(struct sockaddr *nam)
2600 {
2601 	struct sockaddr_in *sin_p;
2602 	struct sockaddr_in6 sin6;
2603 
2604 	/*
2605 	 * Save original sockaddr_in6 addr and convert it
2606 	 * to sockaddr_in.
2607 	 */
2608 	sin6 = *(struct sockaddr_in6 *)nam;
2609 	sin_p = (struct sockaddr_in *)nam;
2610 	in6_sin6_2_sin(sin_p, &sin6);
2611 }
2612 
2613 /* Convert sockaddr_in into sockaddr_in6 in v4 mapped addr format. */
2614 void
2615 in6_sin_2_v4mapsin6_in_sock(struct sockaddr **nam)
2616 {
2617 	struct sockaddr_in *sin_p;
2618 	struct sockaddr_in6 *sin6_p;
2619 
2620 	sin6_p = malloc(sizeof(*sin6_p), M_SONAME, M_WAITOK);
2621 	sin_p = (struct sockaddr_in *)*nam;
2622 	in6_sin_2_v4mapsin6(sin_p, sin6_p);
2623 	free(*nam, M_SONAME);
2624 	*nam = (struct sockaddr *)sin6_p;
2625 }
2626